From case at twd.net Tue Apr 2 18:36:52 2002 From: case at twd.net (Jeremy Rusnak) Date: Tue, 2 Apr 2002 15:36:52 -0800 Subject: [m_b_users] [PATCH] local buffering - revised In-Reply-To: <20020314204256.GB10185@tripie> Message-ID: Hi all, I've downloaded the local buffering patch (woot) and I'm trying to apply it, but I'm having some problems. I'm using: patch -p0 < mod_backhand-1.2.1.localbuf.patch Things seem to go along okay until: patching file `mod_backhand-1.2.1/Makefile.in' patching file `mod_backhand-1.2.1/Makefile.tmpl' patching file `mod_backhand-1.2.1/debug.h' patching file `mod_backhand-1.2.1/localbuf.c' patching file `mod_backhand-1.2.1/localbuf.h' patching file `mod_backhand-1.2.1/mod_backhand.c' Hunk #3 succeeded at 153 with fuzz 2. Hunk #5 FAILED at 588. Hunk #6 FAILED at 674. 2 out of 8 hunks FAILED -- saving rejects to mod_backhand-1.2.1/mod_backhand.c.rej patching file `mod_backhand-1.2.1/mod_backhand.h' I'm admittedly a patch newbie...I don't use it that often. Any suggestions folks? Thanks very much, Jeremy ___________________________________________ Jeremy 'Case' Rusnak President & Founder Case's Ladder, Inc. http://www.igl.net/ ___________________________________________ -----Original Message----- From: backhand-users-admin at lists.backhand.org [mailto:backhand-users-admin at lists.backhand.org]On Behalf Of Tomas Styblo Sent: Thursday, March 14, 2002 12:43 PM To: backhand-users at lists.backhand.org Subject: [m_b_users] [PATCH] local buffering - revised Hello. I prepared the patch for original mod_backhand 2.1.2. I'd like to point out that: a) local buffering is used for POST data even if local buffering is disabled in httpd.conf => Tt's IMHO in almost every situation better then previous implementation, except when size of the POST data exceeds LocalBufferLimitSize (default: 8 MB). In that situation, the data is stored nowhere => it is relayed directly, the backend connection is not fetched from the pool and only one try to establish the connection is made (because we don't store the data anywhere and we obviously can't read it more then once). b) default values are 256 KB for LocalBufferLimitMem and 8 MB for LocalBufferLimitSize c) There is a potential pitfall, which is not really a problem when local buffering is disabled, or when it's used only in situations where it makes sense (on local networks). The problem is that we read all data from the backend and when we are done, then we start to send the data to the client. If the total time to read the data from the backend exceeds many seconds, than the client may close the connection, before we have a chance to send any of the data. Please don't hesitate to ask any questions. -- Tomas Styblo Programmer Czech On Line, a.s. U nakladoveho nadrazi 8, 130 00 Praha 3, Czech Republic From case at twd.net Tue Apr 2 18:36:51 2002 From: case at twd.net (Jeremy Rusnak) Date: Tue, 2 Apr 2002 15:36:51 -0800 Subject: [m_b_users] Backhand and mod_gzip In-Reply-To: <04E1CB9C89835049BF9DDDF4720A4F104BCCB5@mail.rottentomatoes.com> Message-ID: Stephen, I'm not sure if you got any responses off the list but I was curious about this as well. We're using mod_gzip on the tier 2 machines but not our backhand machines. I kind of assumed that the backhand box would simply proxy the request and not actually compress it using mod_gzip, but I haven't verified this. It would be nice to let the backhand box do mod_gzip to take even more off the application servers on tier two. Jeremy -----Original Message----- From: backhand-users-admin at lists.backhand.org [mailto:backhand-users-admin at lists.backhand.org]On Behalf Of Stephen Wang Sent: Thursday, March 28, 2002 1:31 PM To: 'backhand-users at lists.backhand.org' Subject: [m_b_users] Backhand and mod_gzip Apologies if this was asked before: I have a tiered system (i.e. 1 backhand-enabled head-end server pulling from a pool of many backhand-enabled slave servers). Which servers should be running mod_gzip? Is it redundant to have mod_gzip running on the slave servers if it's already running on the head-end server? Does the load order of the mods in my httpd.conf file make a difference? Ideally, I'd like to turn of mod_gzip on the slave servers and just have it on on the head-end server only if I can guarantee that the outgoing pages from my head-end server are properly compressed. Thanks in advance! ------------------------------------------ Stephen Wang stephen.wang at rottentomatoes.com ------------------------------------------ Stephen's FRESH Pick: MONSOON WEDDING http://www.rottentomatoes.com/movie-1111862 From stephen.wang at rottentomatoes.com Tue Apr 2 18:52:13 2002 From: stephen.wang at rottentomatoes.com (Stephen Wang) Date: Tue, 2 Apr 2002 15:52:13 -0800 Subject: [m_b_users] Backhand and mod_gzip Message-ID: <04E1CB9C89835049BF9DDDF4720A4F104BCCCF@mail.rottentomatoes.com> Jeremy, > We're using mod_gzip on the tier 2 machines but not our > backhand machines. I kind of assumed that the backhand box > would simply proxy the request and not actually compress it > using mod_gzip, but I haven't verified this. > It would be nice to let the backhand box do mod_gzip to take > even more off the application servers on tier two. This is exactly what I was thinking. I understand that mod_gzip probably has a pretty low overhead in general, but anything to take load off of our tier 2 machines which are also pumping out some pretty heavy DB queries would help. I haven't received any replies yet -- Theo? Anyone? Same question as before: Is it redundant to have mod_gzip running on the slave servers if it's already running on the head-end server (in a tiered backhand setup)? Does the load order of the mods in my httpd.conf file make a difference? ------------------------------------------ Stephen Wang stephen.wang at rottentomatoes.com ------------------------------------------ Stephen's FRESH Pick: THE ROOKIE http://www.rottentomatoes.com/m/TheRookie-1112954 From tomas.styblo at jet2web.cz Tue Apr 2 18:58:24 2002 From: tomas.styblo at jet2web.cz (Tomas Styblo) Date: Wed, 3 Apr 2002 01:58:24 +0200 Subject: [m_b_users] [PATCH] local buffering - revised In-Reply-To: References: <20020314204256.GB10185@tripie> Message-ID: <20020402235824.GA9166@tripie> > I've downloaded the local buffering patch (woot) and I'm trying to apply it, > but I'm having some problems. Try it this way: tar xvzf mod_backhand-1.2.1.tar.gz cd mod_backhand-1.2.1 cat mod_backhand-1.2.1.localbuf.patch | patch -p1 Be sure to use original unmodified tarball. It works for me using GNU patch 2.5.4. If it still would not work for you, then I can mail the patch to you to ensure you've got an uncorrupted version. -- Tomas Styblo Programmer Czech On Line, a.s. U nakladoveho nadrazi 8, 130 00 Praha 3, Czech Republic From case at twd.net Tue Apr 2 19:28:08 2002 From: case at twd.net (Jeremy Rusnak) Date: Tue, 2 Apr 2002 16:28:08 -0800 Subject: [m_b_users] Backhand and mod_gzip In-Reply-To: <04E1CB9C89835049BF9DDDF4720A4F104BCCCF@mail.rottentomatoes.com> Message-ID: Stephen, This should be fairly easy to test. We just need to see if backhand still parses through gzip when it proxies. I should be able to test this tomorrow and report back to the list, if no one else has tried it. I have been building backhand in static with Apache, so I'm not sure if I will be able to help with the question regarding the order of the modules. Jeremy -----Original Message----- From: backhand-users-admin at lists.backhand.org [mailto:backhand-users-admin at lists.backhand.org]On Behalf Of Stephen Wang Sent: Tuesday, April 02, 2002 3:52 PM To: 'backhand-users at lists.backhand.org' Subject: RE: [m_b_users] Backhand and mod_gzip Jeremy, > We're using mod_gzip on the tier 2 machines but not our > backhand machines. I kind of assumed that the backhand box > would simply proxy the request and not actually compress it > using mod_gzip, but I haven't verified this. > It would be nice to let the backhand box do mod_gzip to take > even more off the application servers on tier two. This is exactly what I was thinking. I understand that mod_gzip probably has a pretty low overhead in general, but anything to take load off of our tier 2 machines which are also pumping out some pretty heavy DB queries would help. I haven't received any replies yet -- Theo? Anyone? Same question as before: Is it redundant to have mod_gzip running on the slave servers if it's already running on the head-end server (in a tiered backhand setup)? Does the load order of the mods in my httpd.conf file make a difference? ------------------------------------------ Stephen Wang stephen.wang at rottentomatoes.com ------------------------------------------ Stephen's FRESH Pick: THE ROOKIE http://www.rottentomatoes.com/m/TheRookie-1112954 _______________________________________________ backhand-users mailing list backhand-users at lists.backhand.org http://lists.backhand.org/mailman/listinfo/backhand-users From case at twd.net Tue Apr 2 19:34:09 2002 From: case at twd.net (Jeremy Rusnak) Date: Tue, 2 Apr 2002 16:34:09 -0800 Subject: [m_b_users] [PATCH] local buffering - revised In-Reply-To: <20020402235824.GA9166@tripie> Message-ID: Thomas, Could you E-mail me the patch? I'm still getting the same result here. It must have been messed up by my mailer. I appreciate it, I'm working on a new server deployment and think this patch would be a great performance boost. Jeremy ___________________________________________ Jeremy 'Case' Rusnak President & Founder Case's Ladder, Inc. http://www.igl.net/ ___________________________________________ -----Original Message----- From: backhand-users-admin at lists.backhand.org [mailto:backhand-users-admin at lists.backhand.org]On Behalf Of Tomas Styblo Sent: Tuesday, April 02, 2002 3:58 PM To: backhand-users at lists.backhand.org Subject: Re: [m_b_users] [PATCH] local buffering - revised > I've downloaded the local buffering patch (woot) and I'm trying to apply it, > but I'm having some problems. Try it this way: tar xvzf mod_backhand-1.2.1.tar.gz cd mod_backhand-1.2.1 cat mod_backhand-1.2.1.localbuf.patch | patch -p1 Be sure to use original unmodified tarball. It works for me using GNU patch 2.5.4. If it still would not work for you, then I can mail the patch to you to ensure you've got an uncorrupted version. -- Tomas Styblo Programmer Czech On Line, a.s. U nakladoveho nadrazi 8, 130 00 Praha 3, Czech Republic _______________________________________________ backhand-users mailing list backhand-users at lists.backhand.org http://lists.backhand.org/mailman/listinfo/backhand-users From case at twd.net Tue Apr 2 20:30:59 2002 From: case at twd.net (Jeremy Rusnak) Date: Tue, 2 Apr 2002 17:30:59 -0800 Subject: [m_b_users] Backhand and mod_gzip testing In-Reply-To: Message-ID: Hi all, I had a chance to test using mod_gzip in tandem with backhand this afternoon. I almost melted my server but I'm happy to say that it seems to work quite well. My limiting factor was that I tested using an older dual P2 450 machine that sits in front of three dual P3 900 servers. The P2 keeps up just fine when it's just proxying or redirecting, but when I turned on gzip after 10 minutes or so it spiraled out of control load wise (it's been stripped to 256 megs of RAM). FYI, this was tested in a production environment that was getting around 200 requests/second. I think with a more powerful machine there wouldn't have been any problems at all. I think in particular with mod_gzip using negotiate "on" having the front line machine doing the compression would be a great performance boost, since you can hit the gzip cache on a central point instead of having n copies on tier two servers. Jeremy ___________________________________________ Jeremy 'Case' Rusnak President & Founder Case's Ladder, Inc. http://www.igl.net/ ___________________________________________ From case at twd.net Tue Apr 2 21:11:10 2002 From: case at twd.net (Jeremy Rusnak) Date: Tue, 2 Apr 2002 18:11:10 -0800 Subject: [m_b_users] Some direction on a new design In-Reply-To: Message-ID: Hi all, I know I've been sending a good bit to the list today, I've been working on a new project that's going to deploy in a few weeks. Currently, the servers under my command are on their own subnet and all have public IPs. Until the last three months or so, we were "load balancing" by moving specific functions to different machines (for example, search, cgi, perl). Now that I have been playing with backhand, I've been working on putting backhand machines in front and sending requests to the applications servers via a redirect (which works great). With the shortage of IP addresses floating around out there and for "greater security", corporate IT wants to reduce the number of IPs they are allocating. What I'd like to do is put a couple uber-backhand boxes up with public IPs and then put the applications machines behind them on a private network (192.168.x.x). This way we can add machines to our hearts content. Is there anyone using backhand in such a configuration on the list who would like to share their experiences? I'm trying to figure out how beefy I should make the backhand boxes if I'm going to be proxying through them, as well as any thoughts on fault tolerance. Any real world examples I can look at? The mailing list archives don't appear to be searchable, I did poke around in there a bit. Thanks, Jeremy From jesus at omniti.com Tue Apr 2 22:24:11 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Tue, 2 Apr 2002 22:24:11 -0500 Subject: [m_b_users] Backhand and mod_gzip In-Reply-To: Message-ID: <4436B4FE-46B2-11D6-A786-00039358205C@omniti.com> On Tuesday, April 2, 2002, at 06:36 PM, Jeremy Rusnak wrote: > It would be nice to let the backhand box do mod_gzip to take > even more off the application servers on tier two. mod_backhand should be able to work fine with mod_gzip... You should be able to compress on your tier one machines. > Apologies if this was asked before: > I have a tiered system (i.e. 1 backhand-enabled head-end server pulling > from > a pool of many backhand-enabled slave servers). Which servers should be > running mod_gzip? Is it redundant to have mod_gzip running on the slave > servers if it's already running on the head-end server? Does the load > order > of the mods in my httpd.conf file make a difference? Yes. It matters quite a bit. But, it should only matter relative to mod_backhand. So move mod_gzip ahead of mod_backhand and try it that way and if it fails, move it after mod_backhand. I find that trial and error is 10 times faster than trying to figure out how all the modules are interacting. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From jesus at omniti.com Tue Apr 2 22:27:45 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Tue, 2 Apr 2002 22:27:45 -0500 Subject: [m_b_users] Backhand and mod_gzip In-Reply-To: Message-ID: On Tuesday, April 2, 2002, at 07:28 PM, Jeremy Rusnak wrote: > I have been building backhand in static with Apache, so I'm not sure > if I will be able to help with the question regarding the order of > the modules. Compiling statically doesn't effect the module order. It isn't the "load" order that concerns us, it is the invokation order. This is decided by the AddModule section in httpd.conf. Look at the default Apache httpd.conf file for reference. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From jesus at omniti.com Tue Apr 2 22:54:25 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Tue, 2 Apr 2002 22:54:25 -0500 Subject: [m_b_users] Some direction on a new design In-Reply-To: Message-ID: <7D72017A-46B6-11D6-A786-00039358205C@omniti.com> On Tuesday, April 2, 2002, at 09:11 PM, Jeremy Rusnak wrote: > Is there anyone using backhand in such a configuration on the > list who would like to share their experiences? I'm trying to > figure out how beefy I should make the backhand boxes if I'm > going to be proxying through them, as well as any thoughts on > fault tolerance. I have mod_backhand running in one, two and three tier architectures. (three being a hardware HA device as the first tier). > Any real world examples I can look at? The mailing list > archives don't appear to be searchable, I did poke around in > there a bit. I am a big advocate of hardware HA devices, I just don't buy into them being the end-all be-all of load balancing. I have run backhand in a two-tier architecture as you have described both with and without a hardware LB/HA device in front (BIG/ip, Arrowpoint and ServerIron in my personal experience). If you are looking at a completely software solution there are many many choices. I don't know how I feel about using wackamole for apache servers (though I've tested it and verified it "works"). I like wackamole running on much more network-oriented devices like firewalls. If you set up two FreeBSD boxes and run a "load balancing" TCP forwarding ruleset on them (to rewrite TCP sessions back to a set of first-tier mod_backhand machines) I think you will be happy with the stability and HA that it provides. In a similar set up, I have two FreeBSD FW/routers running wackamole and the the failover with wackamole is reliable and seamless. I suck at ASCII diagrams, so I won't even go down that road :-) There are a lot of serious design decisions that need to be made when designing a large enterprise oriented architecture. If this is your 1st or 5th, you are still likey to make some decisions that lead to pockets of bad implementation. The key is to make sure that your overall architecture is flexible enough to allow you to reimplement these poorly constructed parts without affecting the global design. Most often these "bad implementations" are due to improper specifications -- or so we engineers like to tell ourselves :-D Another thing to seriously consider is how you plan on handling your logging infrastructure over your seemingly heterogeneous cluster. For that use mod_log_spread, it will simply do the job the right way with a scalable approach. Visit the www.backhand.org mainpage, you might find everything you are looking for. As I said above, I am an advocate of hardware HA devices. You can pick up a used ServerIron for less than $2k now and they are just damn reliable. If you are using mod_backhand on the front tier of web servers anyway, you need not be concerned with how "well" if (HA device) "load balances." -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From alexandra.schwebs at aon.at Wed Apr 3 16:39:44 2002 From: alexandra.schwebs at aon.at (Alexandra Schwebs) Date: Wed, 3 Apr 2002 23:39:44 +0200 Subject: [m_b_users] SetHandler Message-ID: hi group, i have set up backhand on two machines where hp-application server is installed. everything works (backhand and hp-as separately), and now i want backhand to handle also requests for http://host/hpas but i don?t know how to set it up. hp-application server needs following entry in httpd.conf: SetHandler sapphire just adding backhand directives in directive seems not to work. Although i use removeSelf or byRandom only the apache which i called from my browser is serving the request, never the other one, like with other directories etc. maybe because of the SetHandler sapphire directive?? is it not allowed to use it together with backhand directives? can someone help me please? best regards, Alexandra Schwebs alexandra.schwebs at aon.at From case at twd.net Wed Apr 3 17:50:29 2002 From: case at twd.net (Jeremy Rusnak) Date: Wed, 3 Apr 2002 14:50:29 -0800 Subject: [m_b_users] Some direction on a new design In-Reply-To: <7D72017A-46B6-11D6-A786-00039358205C@omniti.com> Message-ID: Theo, Thanks for the feedback. This is very helpful. Hopefully I'll have to time to put together some online documentation when all is said and done that others can read. Jeremy -----Original Message----- From: backhand-users-admin at lists.backhand.org [mailto:backhand-users-admin at lists.backhand.org]On Behalf Of Theo Schlossnagle Sent: Tuesday, April 02, 2002 7:54 PM To: backhand-users at lists.backhand.org Subject: Re: [m_b_users] Some direction on a new design On Tuesday, April 2, 2002, at 09:11 PM, Jeremy Rusnak wrote: > Is there anyone using backhand in such a configuration on the > list who would like to share their experiences? I'm trying to > figure out how beefy I should make the backhand boxes if I'm > going to be proxying through them, as well as any thoughts on > fault tolerance. I have mod_backhand running in one, two and three tier architectures. (three being a hardware HA device as the first tier). > Any real world examples I can look at? The mailing list > archives don't appear to be searchable, I did poke around in > there a bit. I am a big advocate of hardware HA devices, I just don't buy into them being the end-all be-all of load balancing. I have run backhand in a two-tier architecture as you have described both with and without a hardware LB/HA device in front (BIG/ip, Arrowpoint and ServerIron in my personal experience). If you are looking at a completely software solution there are many many choices. I don't know how I feel about using wackamole for apache servers (though I've tested it and verified it "works"). I like wackamole running on much more network-oriented devices like firewalls. If you set up two FreeBSD boxes and run a "load balancing" TCP forwarding ruleset on them (to rewrite TCP sessions back to a set of first-tier mod_backhand machines) I think you will be happy with the stability and HA that it provides. In a similar set up, I have two FreeBSD FW/routers running wackamole and the the failover with wackamole is reliable and seamless. I suck at ASCII diagrams, so I won't even go down that road :-) There are a lot of serious design decisions that need to be made when designing a large enterprise oriented architecture. If this is your 1st or 5th, you are still likey to make some decisions that lead to pockets of bad implementation. The key is to make sure that your overall architecture is flexible enough to allow you to reimplement these poorly constructed parts without affecting the global design. Most often these "bad implementations" are due to improper specifications -- or so we engineers like to tell ourselves :-D Another thing to seriously consider is how you plan on handling your logging infrastructure over your seemingly heterogeneous cluster. For that use mod_log_spread, it will simply do the job the right way with a scalable approach. Visit the www.backhand.org mainpage, you might find everything you are looking for. As I said above, I am an advocate of hardware HA devices. You can pick up a used ServerIron for less than $2k now and they are just damn reliable. If you are using mod_backhand on the front tier of web servers anyway, you need not be concerned with how "well" if (HA device) "load balances." -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 _______________________________________________ backhand-users mailing list backhand-users at lists.backhand.org http://lists.backhand.org/mailman/listinfo/backhand-users From mattcauth at yahoo.com Thu Apr 4 20:23:24 2002 From: mattcauth at yahoo.com (Matt C.) Date: Thu, 4 Apr 2002 17:23:24 -0800 (PST) Subject: [m_b_users] Some direction on a new design In-Reply-To: <7D72017A-46B6-11D6-A786-00039358205C@omniti.com> Message-ID: <20020405012324.54969.qmail@web14002.mail.yahoo.com> I've said it once before, I'll say it again. This list, although tailored to a specific thing, has proven to be extremely educational in many ways. The list alone is worth the price of admission (yeah yeah I know it's open source, but it's a metaphor :), then you throw in backhand itself and things get real interesting. Well done everyone... Regards, Matt --- Theo Schlossnagle wrote: > On Tuesday, April 2, 2002, at 09:11 PM, Jeremy Rusnak wrote: > > Is there anyone using backhand in such a configuration on the > > list who would like to share their experiences? I'm trying to > > figure out how beefy I should make the backhand boxes if I'm > > going to be proxying through them, as well as any thoughts on > > fault tolerance. > > I have mod_backhand running in one, two and three tier architectures. > (three being a hardware HA device as the first tier). > > > Any real world examples I can look at? The mailing list > > archives don't appear to be searchable, I did poke around in > > there a bit. > > I am a big advocate of hardware HA devices, I just don't buy into them > being the end-all be-all of load balancing. I have run backhand in a > two-tier architecture as you have described both with and without a > hardware LB/HA device in front (BIG/ip, Arrowpoint and ServerIron in my > personal experience). > > If you are looking at a completely software solution there are many many > choices. I don't know how I feel about using wackamole for apache > servers (though I've tested it and verified it "works"). I like > wackamole running on much more network-oriented devices like firewalls. > If you set up two FreeBSD boxes and run a "load balancing" TCP > forwarding ruleset on them (to rewrite TCP sessions back to a set of > first-tier mod_backhand machines) I think you will be happy with the > stability and HA that it provides. In a similar set up, I have two > FreeBSD FW/routers running wackamole and the the failover with wackamole > is reliable and seamless. > > I suck at ASCII diagrams, so I won't even go down that road :-) > > There are a lot of serious design decisions that need to be made when > designing a large enterprise oriented architecture. If this is your 1st > or 5th, you are still likey to make some decisions that lead to pockets > of bad implementation. The key is to make sure that your overall > architecture is flexible enough to allow you to reimplement these poorly > constructed parts without affecting the global design. Most often these > "bad implementations" are due to improper specifications -- or so we > engineers like to tell ourselves :-D > > Another thing to seriously consider is how you plan on handling your > logging infrastructure over your seemingly heterogeneous cluster. For > that use mod_log_spread, it will simply do the job the right way with a > scalable approach. Visit the www.backhand.org mainpage, you might find > everything you are looking for. > > As I said above, I am an advocate of hardware HA devices. You can pick > up a used ServerIron for less than $2k now and they are just damn > reliable. If you are using mod_backhand on the front tier of web > servers anyway, you need not be concerned with how "well" if (HA device) > "load balances." > > -- > Theo Schlossnagle > Principal Consultant > OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ > Phone: +1 301 776 6376 Fax: +1 410 880 4879 > 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 > 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 > > > _______________________________________________ > backhand-users mailing list > backhand-users at lists.backhand.org > http://lists.backhand.org/mailman/listinfo/backhand-users __________________________________________________ Do You Yahoo!? Yahoo! Tax Center - online filing with TurboTax http://taxes.yahoo.com/ From jesus at omniti.com Fri Apr 5 11:56:27 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Fri, 05 Apr 2002 11:56:27 -0500 Subject: [m_b_users] Porblem with my session_id and zope References: <5.1.0.14.0.20020328102010.00a89a90@mail.digitome.com> Message-ID: <3CADD73B.9010501@omniti.com> Derek Higgins wrote: > I was having a lot of trouble with the bySession candidacy function > lately when no matter what I did values I set for the identifier seemed > to be getting ignored. [snip] > the apache servers that the requests are being sent to are inturn > proxying the requests to > a zope server on a different port number and its on the zope server that > the cookies are being set > RESPONSE.setCookie("mcid", "ac110a12") > [ ... snip ... ] > Which meant that the decode loop was getting a " as the first character > to decode as a hex digit and it > all went wronge from there > > This means I can get the session_id to work if I replace the bySession > configuration in my http.conf with > Backhand bySession mcid=" > > I have three questions > Does this mean that the zope web server sets cookies different to the > rest of the web servers in the world? So it would seem > Should this be addressed in the code for bySession? No > The README.bySession file talks about a ap_log_error statement that can > be uncommented for debugging, where is it? It was probably commented out permanently (deleted). It might be in the CVS history. Most likely you will want to add your own ap_log_error line if you are actually planning on debugging it. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From stephen.wang at rottentomatoes.com Fri Apr 5 20:03:44 2002 From: stephen.wang at rottentomatoes.com (Stephen Wang) Date: Fri, 5 Apr 2002 17:03:44 -0800 Subject: [m_b_users] byBusyChildren and # of servers count inquiries Message-ID: <04E1CB9C89835049BF9DDDF4720A4F104BCCDA@mail.rottentomatoes.com> Two inquiries: * Looking at the 1.2.1 Changelog, it appears that there was previously a problem with byBusyChildren calling byLoad that was apparently fixed. I'm browsing the 1.2.1 builtin.h:52 and it looks like to me that it's still calling byLoad. The CVS reflects that this particular file hasn't been modified since before 1.2.1 as well. Am I completely offbase or is byBusyChildren still "byLoad" in the current CVS? * I'm continuing to get "# ready servers/# total servers" count of "0/0" for all of my servers on the stat page. I've tried recompiling apache as well (since I modded the upper limit on the processes count), but to no avail. I'm correct in assuming that these stats are used when using the byBusyChildren criteria, right? Is there something else that I need to do in order to get the numserver count up-and-running again? Advice would be greatly appreciated. ------------------------------------------ Stephen Wang stephen.wang at rottentomatoes.com ------------------------------------------ Stephen's FRESH Pick: THE ROOKIE http://www.rottentomatoes.com/m/TheRookie-1112954 From case at twd.net Fri Apr 5 21:10:40 2002 From: case at twd.net (Jeremy Rusnak) Date: Fri, 5 Apr 2002 18:10:40 -0800 Subject: [m_b_users] byBusyChildren and # of servers count inquiries In-Reply-To: <04E1CB9C89835049BF9DDDF4720A4F104BCCDA@mail.rottentomatoes.com> Message-ID: Stephen, On the second question, if you kill -HUP Apache and restart it it will start reporting the correct number of servers. This is addressed online in the FAQ I believe. Here's a couple lines I have used to automate this process: /home/httpd/httpd -f /home/httpd/conf/httpd.conf sleep 2 kill -HUP `cat /home/httpd/logs/httpd.pid` sleep 2 renice -20 `tail /home/httpd/logs/error_log|grep 'spawning'|grep PID|awk '{print substr($11,0,length($11)-1)}'` This addresses both the concern of seeing the right number of servers reported on the mod_backhand display, as well as renices the moderator process (as recommended online in the performance area). I've noticed if you're spawning a large number of children when you start (say, 300) you might have to increase the sleep amount. Hope this is of some help! Jeremy -----Original Message----- From: backhand-users-admin at lists.backhand.org [mailto:backhand-users-admin at lists.backhand.org]On Behalf Of Stephen Wang Sent: Friday, April 05, 2002 5:04 PM To: 'backhand-users at lists.backhand.org' Subject: [m_b_users] byBusyChildren and # of servers count inquiries Two inquiries: * Looking at the 1.2.1 Changelog, it appears that there was previously a problem with byBusyChildren calling byLoad that was apparently fixed. I'm browsing the 1.2.1 builtin.h:52 and it looks like to me that it's still calling byLoad. The CVS reflects that this particular file hasn't been modified since before 1.2.1 as well. Am I completely offbase or is byBusyChildren still "byLoad" in the current CVS? * I'm continuing to get "# ready servers/# total servers" count of "0/0" for all of my servers on the stat page. I've tried recompiling apache as well (since I modded the upper limit on the processes count), but to no avail. I'm correct in assuming that these stats are used when using the byBusyChildren criteria, right? Is there something else that I need to do in order to get the numserver count up-and-running again? Advice would be greatly appreciated. ------------------------------------------ Stephen Wang stephen.wang at rottentomatoes.com ------------------------------------------ Stephen's FRESH Pick: THE ROOKIE http://www.rottentomatoes.com/m/TheRookie-1112954 _______________________________________________ backhand-users mailing list backhand-users at lists.backhand.org http://lists.backhand.org/mailman/listinfo/backhand-users From lraynaud at Barneki.com Sun Apr 7 10:51:47 2002 From: lraynaud at Barneki.com (Lionel Raynaud) Date: Sun, 07 Apr 2002 10:51:47 -0400 Subject: [m_b_users] Basic but strange Message-ID: <3CB05D03.4050109@Barneki.com> We are actually building a two tiers solution using mod_backhand. In implementing this solution we found a behavior that sounds not clearly explainable (at least to us ;-). On the frontend servers the use of the following simple configuration: # # This may also be "None", "All", or any combination of "Indexes", # "Includes", "FollowSymLinks", "ExecCGI", or "MultiViews". # # Note that "MultiViews" must be named *explicitly* --- "Options All" # doesn't give it to you. # Options Indexes FollowSymLinks MultiViews AllowOverride None Backhand byAge Backhand byRandom .... MulticastStats 192.168.1.255:4445 AcceptStats 192.168.1.0/24 ... Either the default apache page is displayed properly, either a HTTP negative answer is received as: - "Not Acceptable" An appropriate representation of the requested resource /index.html could not be found on this server.... - "Index of /" ... Note that if the Apache "Option Indexes" is not allowed, the HTTP negative answer is becoming "Forbiden (403) - You don't have permission to access / on this server..." Interesting too: - this behaviour is the same independently of the front-end servers queried, two in our tests. - This error does not appear as often IF we replace put byLoad instead of byRandom. ----------------------------------------------------------------- Here a quick description of our topology: Frontend: 192.168.1.0/24 - backhend: 192.168.2.0/24. After a proper installation of apache-1.3.24 + mod_backhand-1.2.1 on multiple hosts (backend and/or frontend) (Solaris 8, 7, FreeBSD, Linux) we validated quickly the solution simply by connecting a browser to each apache servers on their :/backhand/. Outside of the bug with 0/0 servers available,the multicasting of information is working perfectly. Afterwards, we focused on the front-end servers only and make our first test as described above. Just to clarify questions on this test, this is exactly what has been done: - We stopped the backend servers - Useless at this point - - Changed the front-end servers httpd.conf config to the one described below onthe 2 front-end: 192.168.1.10 & 192.168.1.20 - Restarted the front-end - Validated that :/backhand/ is displaying the correct 2 servers - Use a browser on 192.168.1.101 on the same segment with No local browsing cache (netscape 6.2). Thx Lionel From mike at vpi.net Thu Apr 11 19:41:06 2002 From: mike at vpi.net (Mike Jimenez) Date: Thu, 11 Apr 2002 16:41:06 -0700 Subject: [m_b_users] New to backhand Message-ID: <000a01c1e1b2$599b68f0$208d3a40@kludge> Hi, I am looking for a load balancing solution, However I am looking to balance the load for the entire website rather then just a directory ex: cgi-bin Is mod backhand the solution? I already loaded the mod on one server but only want to run it on 2 servers rather then 10 or more. My other question is how would I configure it for just 2 servers using the following example. Assuming that my broadcast IP is 10.0.0.10 and my other server is 10.0.0.11 Is the following a valid config? # This directorty must be readable and writable by euid of apache (nobody) # The mod_backhand-Arriba and the AF_UNIX domain files are stored in here. UnixSocketDir /var/backhand/backhand # This would be the way to do ethernet broadcast # MulticastStats 128.220.221.255:4445 # but we choose to use the following IP multicast address with a TTL of 1 MulticastStats 10.0.0.10:4445,1 # We want to accept resource information originating from any IP on our # network. AcceptStats 10.0.0.11 Thanks in advance Mike From case at twd.net Fri Apr 12 02:49:41 2002 From: case at twd.net (Jeremy Rusnak) Date: Thu, 11 Apr 2002 23:49:41 -0700 Subject: [m_b_users] New to backhand In-Reply-To: <000a01c1e1b2$599b68f0$208d3a40@kludge> Message-ID: Mike, Welcome! I'll save the configuration questions for someone else since I'm fairly new to backhand and don't want to steer you wrong. The only trouble you may run into with backhand with only two servers is that it seems designed to handle larger pools. Using some of the standard candidacy funtions (such as byBusyChildren or byLoad) on only two machines can cause some problems if you don't tweak things correctly. Since you only have two choices, the problem I had was that all requests were either going to one machine or the other at any given time. I'd see one machine with 50 busy clients and the other with 10, then it would flip flop. Backhand really starts to work well once you get beyond two servers, so you can eliminate some of them from the pool using byLogWindow. You can also tweak the settings a bit to keep requests on the local machine more often than forwarding them. If anyone else has any good suggestions on the above scenario, I'd love to hear them as well. Maybe there is a much smoother way to implement across only two servers, but I really have not found it yet. If you know for sure that you will always only need to balance your entire server, you might look at something like LVS (www.linuxvirtualserver.org). You can get higher throughput using a method like LVS, but you don't have the fine control that you do with backhand. Personally, I prefer backhand since it's easier to administer and I can really fine tune what I want balanced and how. Plus you don't have to deal with kernel patches or modules. Jeremy -----Original Message----- From: backhand-users-admin at lists.backhand.org [mailto:backhand-users-admin at lists.backhand.org]On Behalf Of Mike Jimenez Sent: Thursday, April 11, 2002 4:41 PM To: backhand-users at lists.backhand.org Subject: [m_b_users] New to backhand Hi, I am looking for a load balancing solution, However I am looking to balance the load for the entire website rather then just a directory ex: cgi-bin Is mod backhand the solution? I already loaded the mod on one server but only want to run it on 2 servers rather then 10 or more. My other question is how would I configure it for just 2 servers using the following example. Assuming that my broadcast IP is 10.0.0.10 and my other server is 10.0.0.11 Mike From korte at hk-net.de Fri Apr 12 09:01:13 2002 From: korte at hk-net.de (Martin Korte) Date: Fri, 12 Apr 2002 13:01:13 -0000 Subject: [m_b_users] #ready servers/# total servers 0/0 Message-ID: <20020412150113.1b1ufp2@web2.hk-net.de> I?m using 3 Servers (SuSE 7.3, Apache 1.3.20, modbackhand 1.2.0, RPM?s from CD). The configuration: host.domain.com UnixSocketDir /var/lib/backhand MulticastStats X.X.X.204 X.X.X.255:4445 AcceptStats X.X.X.0/24 SetHandler backhand-handler AllowOverride None Options None Order allow,deny Allow from all Backhand byAge Backhand byRandom Backhand byLogWindow Backhand addSelf Backhand byLoad 2.0 host2.domain.com UnixSocketDir /var/lib/backhand MulticastStats X.X.X.255:4445 AcceptStats X.X.X.0/24 AllowOverride None Options None Order allow,deny Allow from all Backhand byAge Backhand byRandom Backhand byLogWindow Backhand addSelf Backhand byLoad 2.0 host3.domain.com UnixSocketDir /var/lib/backhand MulticastStats X.X.X.255:4445 AcceptStats X.X.X.0/24 AllowOverride None Options None Order allow,deny Allow from all Backhand byAge Backhand byRandom Backhand byLogWindow Backhand addSelf Backhand byLoad 2.0 Result from the status-page: Entry Hostname Age Address Total Mem Avail Mem # ready servers/ # total servers ~ms/req [#req] Arriba # CPUs Load/HWM CPU Idle 0 host.domain.com 0 X.X.X.204:80 2110980096 1917641920 0/0 0 [0] 2670969 2 0.000000/32 1.000000 1 host2.domain.com 1 X.X.X.205:80 525914112 441070784 0/0 0 [0] 2552173 1 0.000000/32 1.000000 2 host3.domain.com 1 X.X.X.206:80 393793536 313062592 0/0 0 [0] 1710733 1 0.000000/32 1.000000 why there is #ready servers/# total servers 0/0? any idea? Martin Korte -- MFG Martin Korte, H?ndle & Korte GmbH Webdesign - Internet Consulting - Connectivity Rathausufer 15, D-40213 Duesseldorf Voice: +49-211-6022440, Fax: +49-211-60224422 email: korte at hk-net.de, http://www.hk-net.de From lraynaud at Barneki.com Fri Apr 12 09:10:13 2002 From: lraynaud at Barneki.com (Lionel Raynaud) Date: Fri, 12 Apr 2002 09:10:13 -0400 Subject: [m_b_users] Basic but strange: by Random doesn't work? Theo ? References: <3CB05D03.4050109@Barneki.com> Message-ID: <3CB6DCB5.9030800@Barneki.com> All, Following up on myself on the issue described in the email below: On the same config than below, byRandom appears to work fine on the default apache cgi directory, into which I dit set the backhand cgi test script. - I tried also to use a complete different directory than the default apache HTML directory: Still doesn't work. - I tried to use Files instead of Directory: Doesn't work - Manu other things: Doesn't work ;-) Any suggestion? Thx Lionel Lionel Raynaud wrote: > > We are actually building a two tiers solution using mod_backhand. In > implementing this solution we found a behavior that sounds not clearly > explainable (at least to us ;-). On the frontend servers the use of the > following simple configuration: > > > # > # This may also be "None", "All", or any combination of "Indexes", > # "Includes", "FollowSymLinks", "ExecCGI", or "MultiViews". > # > # Note that "MultiViews" must be named *explicitly* --- "Options All" > # doesn't give it to you. > # > Options Indexes FollowSymLinks MultiViews > AllowOverride None > Backhand byAge > Backhand byRandom > > > > .... > MulticastStats 192.168.1.255:4445 > AcceptStats 192.168.1.0/24 > ... > > > > Either the default apache page is displayed properly, either a HTTP > negative answer is received as: > > - "Not Acceptable" > An appropriate representation of the requested resource /index.html > could not be found on this server.... > > - "Index of /" > ... > > Note that if the Apache "Option Indexes" is not allowed, the HTTP > negative answer is becoming "Forbiden (403) - You don't have permission > to access / on this server..." > > Interesting too: > - this behaviour is the same independently of the front-end servers > queried, two in our tests. > - This error does not appear as often IF we replace put byLoad instead > of byRandom. > > ----------------------------------------------------------------- > > Here a quick description of our topology: > > Frontend: 192.168.1.0/24 - backhend: 192.168.2.0/24. > > After a proper installation of apache-1.3.24 + mod_backhand-1.2.1 on > multiple hosts (backend and/or frontend) (Solaris 8, 7, FreeBSD, Linux) > we validated quickly the solution simply by connecting a browser to each > apache servers on their :/backhand/. Outside of the bug with 0/0 > servers available,the multicasting of information is working perfectly. > > Afterwards, we focused on the front-end servers only and make our first > test as described above. Just to clarify questions on this test, this is > exactly what has been done: > > - We stopped the backend servers - Useless at this point - > - Changed the front-end servers httpd.conf config to the one described > below onthe 2 front-end: 192.168.1.10 & 192.168.1.20 > - Restarted the front-end > - Validated that :/backhand/ is displaying the correct 2 servers > - Use a browser on 192.168.1.101 on the same segment with No local > browsing cache (netscape 6.2). > > Thx > > Lionel > > > _______________________________________________ > backhand-users mailing list > backhand-users at lists.backhand.org > http://lists.backhand.org/mailman/listinfo/backhand-users From mike at vpi.net Fri Apr 12 12:03:05 2002 From: mike at vpi.net (Mike Jimenez) Date: Fri, 12 Apr 2002 09:03:05 -0700 Subject: [m_b_users] New to backhand In-Reply-To: Message-ID: <011001c1e23b$87d2ede0$208d3a40@kludge> Jeremy, Thanks for the info, If there is anyone out there that can help with providing a solution for 2 server configuration with backhand please let me know. Thanks Mike -----Original Message----- From: backhand-users-admin at lists.backhand.org [mailto:backhand-users-admin at lists.backhand.org] On Behalf Of Jeremy Rusnak Sent: Thursday, April 11, 2002 11:50 PM To: backhand-users at lists.backhand.org Subject: RE: [m_b_users] New to backhand Mike, Welcome! I'll save the configuration questions for someone else since I'm fairly new to backhand and don't want to steer you wrong. The only trouble you may run into with backhand with only two servers is that it seems designed to handle larger pools. Using some of the standard candidacy funtions (such as byBusyChildren or byLoad) on only two machines can cause some problems if you don't tweak things correctly. Since you only have two choices, the problem I had was that all requests were either going to one machine or the other at any given time. I'd see one machine with 50 busy clients and the other with 10, then it would flip flop. Backhand really starts to work well once you get beyond two servers, so you can eliminate some of them from the pool using byLogWindow. You can also tweak the settings a bit to keep requests on the local machine more often than forwarding them. If anyone else has any good suggestions on the above scenario, I'd love to hear them as well. Maybe there is a much smoother way to implement across only two servers, but I really have not found it yet. If you know for sure that you will always only need to balance your entire server, you might look at something like LVS (www.linuxvirtualserver.org). You can get higher throughput using a method like LVS, but you don't have the fine control that you do with backhand. Personally, I prefer backhand since it's easier to administer and I can really fine tune what I want balanced and how. Plus you don't have to deal with kernel patches or modules. Jeremy -----Original Message----- From: backhand-users-admin at lists.backhand.org [mailto:backhand-users-admin at lists.backhand.org]On Behalf Of Mike Jimenez Sent: Thursday, April 11, 2002 4:41 PM To: backhand-users at lists.backhand.org Subject: [m_b_users] New to backhand Hi, I am looking for a load balancing solution, However I am looking to balance the load for the entire website rather then just a directory ex: cgi-bin Is mod backhand the solution? I already loaded the mod on one server but only want to run it on 2 servers rather then 10 or more. My other question is how would I configure it for just 2 servers using the following example. Assuming that my broadcast IP is 10.0.0.10 and my other server is 10.0.0.11 Mike _______________________________________________ backhand-users mailing list backhand-users at lists.backhand.org http://lists.backhand.org/mailman/listinfo/backhand-users From jesus at omniti.com Fri Apr 12 12:05:47 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Fri, 12 Apr 2002 12:05:47 -0400 Subject: [m_b_users] New to backhand References: <011001c1e23b$87d2ede0$208d3a40@kludge> Message-ID: <3CB705DB.5000504@omniti.com> Mike Jimenez wrote: >Thanks for the info, If there is anyone out there that can help with >providing a solution for 2 server configuration with backhand please let >me know. > byRandom should work just fine if the machines are homogeneous. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From jesus at omniti.com Fri Apr 12 12:11:23 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Fri, 12 Apr 2002 12:11:23 -0400 Subject: [m_b_users] #ready servers/# total servers 0/0 References: <20020412150113.1b1ufp2@web2.hk-net.de> Message-ID: <3CB7072B.7060508@omniti.com> Martin Korte wrote: > >why there is #ready servers/# total servers 0/0? any idea? > Read the FAQ. http://www.backhand.org/mod_backhand/FAQ.shtml#question13 From mike at vpi.net Fri Apr 12 12:10:42 2002 From: mike at vpi.net (Mike Jimenez) Date: Fri, 12 Apr 2002 09:10:42 -0700 Subject: [m_b_users] New to backhand In-Reply-To: <3CB705DB.5000504@omniti.com> Message-ID: <011301c1e23c$983a3070$208d3a40@kludge> Theo, I really do appreciate the info however I am really new to backhand and I am not sure I understand what you mean. Can you please explain in a little more detail. Thanks Mike -----Original Message----- From: backhand-users-admin at lists.backhand.org [mailto:backhand-users-admin at lists.backhand.org] On Behalf Of Theo Schlossnagle Sent: Friday, April 12, 2002 9:06 AM To: backhand-users at lists.backhand.org Subject: Re: [m_b_users] New to backhand Mike Jimenez wrote: >Thanks for the info, If there is anyone out there that can help with >providing a solution for 2 server configuration with backhand please >let me know. > byRandom should work just fine if the machines are homogeneous. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 _______________________________________________ backhand-users mailing list backhand-users at lists.backhand.org http://lists.backhand.org/mailman/listinfo/backhand-users From jesus at omniti.com Wed Apr 17 18:24:55 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Wed, 17 Apr 2002 18:24:55 -0400 Subject: [m_b_users] Help!!!! Stuck with mod_backhand References: <002c01c1d05d$ac32a260$83e6453f@ayanami> Message-ID: <3CBDF637.7060300@omniti.com> Gabriel Maffia wrote: > I?m getting this error while starting apache with mod_backhand > > [Wed Apr 17 19:13:24 2002] [error] (2)No such file or directory: Child > 12212 fai > led to establish umbilical to moderator! Sounds like a permissions problem. Look through the mailing list archives, it has been discussed quite a bit. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From jesus at omniti.com Wed Apr 17 18:28:34 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Wed, 17 Apr 2002 18:28:34 -0400 Subject: [m_b_users] Domains and backhand References: <004a01c1d027$e312da10$83e6453f@ayanami> Message-ID: <3CBDF712.3060508@omniti.com> Gabriel Maffia wrote: > I have this domain www.foobar.com , and I will be clustering with > two servers. Now, as far as I?ve read, those servers will need to have > different names like www1.foobar.com and www2.foobar.com. Is there a > way to keep the name www.foobar.com and make > backhand redirect to the appropiate server without changing the URL? The only reason you need to name the different is if you need to have them show up with different names in the /backhand/ summary page, or if you plan on doing HTTP redirection instead of proxying. I always name my webserver differently.. That way when you get an internal server error it says which machine and if you have a web application you don't need to go calling gethostname() to get the machines name, it is in the server configuration. Use a virtualhost and set the ServerName to www.foobar.com. Like: ServerName www-s1-1.hostingcompany.com NameVirtualHost myipaddress:80 ServerName www.foobar.com .... -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From sean at chittenden.org Wed Apr 17 19:31:00 2002 From: sean at chittenden.org (Sean Chittenden) Date: Wed, 17 Apr 2002 16:31:00 -0700 Subject: [m_b_users] Another backhand issue In-Reply-To: <004f01c1d065$03f0c6b0$83e6453f@ayanami>; from "gmaffia@technisys.net" on Wed, Mar 20, 2002 at = 08:14:41PM References: <004f01c1d065$03f0c6b0$83e6453f@ayanami> Message-ID: <20020417163100.A32518@ninja1.internal> > In server1 (the one where the dns for www.foobar.com is) I have this configuration: > > > AllowOverride None > Options None > Order allow,deny > Allow from all > Backhand byAge > Backhand byRandom > Backhand byLogWindow > Backhand byLoad > If you've only got two servers, you shouldn't use byLogWindow. Just use byAge 3 instead. What does the bachand handler page say (nifty stats page)? -sc PS, it sounds like you have the right idea. -- Sean Chittenden From sean at chittenden.org Wed Apr 17 20:00:16 2002 From: sean at chittenden.org (Sean Chittenden) Date: Wed, 17 Apr 2002 17:00:16 -0700 Subject: [m_b_users] Another backhand issue In-Reply-To: <006c01c1d06a$b9cd6100$83e6453f@ayanami>; from "gmaffia@technisys.net" on Wed, Mar 20, 2002 at = 08:55:34PM References: <004f01c1d065$03f0c6b0$83e6453f@ayanami> <20020417163100.A32518@ninja1.internal> <006c01c1d06a$b9cd6100$83e6453f@ayanami> Message-ID: <20020417170016.A32765@ninja1.internal> > Now I?m having in ms/req[#req] for server1 : 46[22] and 0[0] for > server2....it seems that server2 isn't being sent anything... > > I don't know if this is useful in any way I think your diagnosis is dead on the money: server1 is getting all of the traffic. Didn't you say that server1 had www.foobar.com pointed to it? This'll only work if both server1 and server2 have www.foobar.com pointed to them (unless you want server1 to do all of the backhanding: akin to a reverse proxy). Requests/sec = number of backhanded requests, not necessarily the number of requests served. -sc -- Sean Chittenden From alvar at a-blast.org Wed Apr 17 21:16:16 2002 From: alvar at a-blast.org (Alvar Freude) Date: Thu, 18 Apr 2002 03:16:16 +0200 Subject: [m_b_users] Virtual Hosts and other ugly things Message-ID: <1608120000.1019092576@gnarzelwicht.delirium-arts.de> Hello, mod_backhand experts, after long time trying to configure mod_backhand for a small cluster with 3 machines and a lot of virtual hosts I give up and ask here ... I've two questions: first how to handle virtual hosts and second how to handle my special situation. Let's start with the last: The scenario is as follows (given by my boss): * a master server, which handles all originating requests this server delivers some virtual hosts directly and forwards some other to the two backend servers * two additional webservers, which handle some virtual hosts with a lot of slow or "CPU intensive" PHP code * (a fourth server is a database only server with idle time) Yes, I know this is stupid, not redundant and throws a lot of performance away. It's not my decision (but perhaps in future it will revoked) ... If someone has an example configuration for this,.I'll be very happy! :-) At the moment we have an Apache on Port 8080 to test the new scenario (on the frontend is an old apache running all virtual hosts). My Configuration is at the moment: # === FRONTEND SERVER === # socket etc ... UnixSocketDir /var/log/apache/backhand # If a server DOESN'T send some stats to others, # apache/backhand DOESN'T start without errors # So, let's send some dummy stats somewhere ... MulticastStats 212.9.188.89 212.9.188.92:4445 # Accept stats from our two backends AcceptStats 212.9.188.93 AcceptStats 212.9.188.92 # activate load balancing for a directory / testing # the same will be placed in some VHost-Settings Backhand byAge Backhand byLoad SetHandler backhand-handler Allow from all # === BACKEND SERVERS === # sockets ... UnixSocketDir /var/log/apache/backhand # send stats to frontend only MulticastStats 212.9.188.89:4445 # take stats from our subnet # without this apache/mod_backhand doesn't start correct AcceptStats 212.9.188.88/29 # ======================= Nevertheless the frontend accepts stats from localhost, in the server overview at /backhand/ shows a red / not ready 212.9.188.89:8443 (frontend with SSL), the second backend (212.9.188.93) and 127.0.0.1:80 (in this order). Why iappears the frontend? And why as third machine localhost and not 212.9.188.92? 212.9.188.92 appears The second part is: how should I write a correct virtual host config? It seems, that the frontend handles connections, even if there are Backhand byAge Backhand byLoad Statements in the -Config (cgi folder at above example). Why? How can i get rid of this? At the frontend it seems to be easy to deliver sone virtual hosts to the backends: # ... Backhand byAge Backhand byLoad # ... This is correct, isn't it? But how can the backends realize which virtual host is accessed? I tryed it with the following, but something goes wrong: # === BACKEND VHOST === Listen 80 NameVirtualHost 212.9.188.92 ServerAlias web.foobar.de # ... # ====================== How should I configure the backend's vhosts, so that apache recognises them? I can't write the name into the VirtualHost Tag, because it resolves to another IP ... Perhaps the whole scenaria can't work? Is it better to create a DNS round Robin for all Domains and to have an private subnet (e.g. on 192.168.1.0/24) for forwarding and all machines should deliver everything? But how Virtual Hosts work then?!? Puuh, a lot of greenhorn questions. I hope you understand them ... Thank you for helping! Ciao Alvar -- // STOP CENSORSHIP! http://www.odem.org/informationsfreiheit/en/ // Blaster: http://www.a-blast.org/ -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 238 bytes Desc: not available Url : http://lists.omniti.com/pipermail/backhand-users/attachments/20020418/c503f188/attachment.bin From lraynaud at Barneki.com Wed Apr 17 21:31:47 2002 From: lraynaud at Barneki.com (Lionel Raynaud) Date: Wed, 17 Apr 2002 21:31:47 -0400 Subject: [m_b_users] byRandom doesn't work with 2 servers References: <3CB05D03.4050109@Barneki.com> <3CB6DCB5.9030800@Barneki.com> Message-ID: <3CBE2203.7010505@Barneki.com> Having seen no answer on this issue we can probably conclude that this is so well known that nobody dare to answer ;-) or that it's a side effect widly accepted without the need for solution?! ??? Lionel Lionel Raynaud wrote: > All, > > Following up on myself on the issue described in the email below: > > On the same config than below, byRandom appears to work fine on the > default apache cgi directory, into which I dit set the backhand cgi test > script. > > - I tried also to use a complete different directory than the default > apache HTML directory: Still doesn't work. > > - I tried to use Files instead of Directory: Doesn't work > > - Manu other things: Doesn't work ;-) > > Any suggestion? > > Thx > > Lionel > > > > Lionel Raynaud wrote: > >> >> We are actually building a two tiers solution using mod_backhand. In >> implementing this solution we found a behavior that sounds not clearly >> explainable (at least to us ;-). On the frontend servers the use of >> the following simple configuration: >> >> >> # >> # This may also be "None", "All", or any combination of "Indexes", >> # "Includes", "FollowSymLinks", "ExecCGI", or "MultiViews". >> # >> # Note that "MultiViews" must be named *explicitly* --- "Options All" >> # doesn't give it to you. >> # >> Options Indexes FollowSymLinks MultiViews >> AllowOverride None >> Backhand byAge >> Backhand byRandom >> >> >> >> .... >> MulticastStats 192.168.1.255:4445 >> AcceptStats 192.168.1.0/24 >> ... >> >> >> >> Either the default apache page is displayed properly, either a HTTP >> negative answer is received as: >> >> - "Not Acceptable" >> An appropriate representation of the requested resource /index.html >> could not be found on this server.... >> >> - "Index of /" >> ... >> >> Note that if the Apache "Option Indexes" is not allowed, the HTTP >> negative answer is becoming "Forbiden (403) - You don't have >> permission to access / on this server..." >> >> Interesting too: >> - this behaviour is the same independently of the front-end servers >> queried, two in our tests. >> - This error does not appear as often IF we replace put byLoad instead >> of byRandom. >> >> ----------------------------------------------------------------- >> >> Here a quick description of our topology: >> >> Frontend: 192.168.1.0/24 - backhend: 192.168.2.0/24. >> >> After a proper installation of apache-1.3.24 + mod_backhand-1.2.1 on >> multiple hosts (backend and/or frontend) (Solaris 8, 7, FreeBSD, >> Linux) we validated quickly the solution simply by connecting a >> browser to each apache servers on their :/backhand/. Outside of >> the bug with 0/0 servers available,the multicasting of information is >> working perfectly. >> >> Afterwards, we focused on the front-end servers only and make our >> first test as described above. Just to clarify questions on this test, >> this is exactly what has been done: >> >> - We stopped the backend servers - Useless at this point - >> - Changed the front-end servers httpd.conf config to the one described >> below onthe 2 front-end: 192.168.1.10 & 192.168.1.20 >> - Restarted the front-end >> - Validated that :/backhand/ is displaying the correct 2 servers >> - Use a browser on 192.168.1.101 on the same segment with No local >> browsing cache (netscape 6.2). >> >> Thx >> >> Lionel >> >> >> _______________________________________________ >> backhand-users mailing list >> backhand-users at lists.backhand.org >> http://lists.backhand.org/mailman/listinfo/backhand-users > > > > > _______________________________________________ > backhand-users mailing list > backhand-users at lists.backhand.org > http://lists.backhand.org/mailman/listinfo/backhand-users From bondu at alphastate.com Wed Apr 17 22:30:55 2002 From: bondu at alphastate.com (Daniel Bondurant) Date: Wed, 17 Apr 2002 19:30:55 -0700 (PDT) Subject: [m_b_users] byRandom doesn't work with 2 servers In-Reply-To: <3CBE2203.7010505@Barneki.com> References: <3CBE2203.7010505@Barneki.com> Message-ID: <4232.66.12.33.98.1019097055.squirrel@www.alphastate.com> I was having a very similar problem. Check the apache logs, if it is trying to access the file /index.html instead of /path/to/www/index.html, than I was having the same problem. Search the mail list archives for similar problems. What was happening for me was apache was segfaulting every few requests. I never came up with a good solution/workaround, and eventually resorted to rebuilding the machine. > > Having seen no answer on this issue we can probably conclude that > this is so well known that nobody dare to answer ;-) or that it's a > side effect widly accepted without the need for solution?! > > ??? > > Lionel > > > > Lionel Raynaud wrote: > >> All, >> >> Following up on myself on the issue described in the email below: >> >> On the same config than below, byRandom appears to work fine on the >> default apache cgi directory, into which I dit set the backhand cgi >> test script. >> >> - I tried also to use a complete different directory than the default >> apache HTML directory: Still doesn't work. >> >> - I tried to use Files instead of Directory: Doesn't work >> >> - Manu other things: Doesn't work ;-) >> >> Any suggestion? >> >> Thx >> >> Lionel >> >> >> >> Lionel Raynaud wrote: >> >>> >>> We are actually building a two tiers solution using mod_backhand. In >>> implementing this solution we found a behavior that sounds not >>> clearly explainable (at least to us ;-). On the frontend servers the >>> use of the following simple configuration: >>> >>> >>> # >>> # This may also be "None", "All", or any combination of "Indexes", # >>> "Includes", "FollowSymLinks", "ExecCGI", or "MultiViews". >>> # >>> # Note that "MultiViews" must be named *explicitly* --- "Options All" >>> # doesn't give it to you. >>> # >>> Options Indexes FollowSymLinks MultiViews >>> AllowOverride None >>> Backhand byAge >>> Backhand byRandom >>> >>> >>> >>> .... >>> MulticastStats 192.168.1.255:4445 >>> AcceptStats 192.168.1.0/24 >>> ... >>> >>> >>> >>> Either the default apache page is displayed properly, either a HTTP >>> negative answer is received as: >>> >>> - "Not Acceptable" >>> An appropriate representation of the requested resource /index.html >>> could not be found on this server.... >>> >>> - "Index of /" >>> ... >>> >>> Note that if the Apache "Option Indexes" is not allowed, the HTTP >>> negative answer is becoming "Forbiden (403) - You don't have >>> permission to access / on this server..." >>> >>> Interesting too: >>> - this behaviour is the same independently of the front-end servers >>> queried, two in our tests. >>> - This error does not appear as often IF we replace put byLoad >>> instead of byRandom. >>> >>> ----------------------------------------------------------------- >>> >>> Here a quick description of our topology: >>> >>> Frontend: 192.168.1.0/24 - backhend: 192.168.2.0/24. >>> >>> After a proper installation of apache-1.3.24 + mod_backhand-1.2.1 on >>> multiple hosts (backend and/or frontend) (Solaris 8, 7, FreeBSD, >>> Linux) we validated quickly the solution simply by connecting a >>> browser to each apache servers on their :/backhand/. Outside of >>> the bug with 0/0 servers available,the multicasting of information is >>> working perfectly. >>> >>> Afterwards, we focused on the front-end servers only and make our >>> first test as described above. Just to clarify questions on this >>> test, this is exactly what has been done: >>> >>> - We stopped the backend servers - Useless at this point - >>> - Changed the front-end servers httpd.conf config to the one >>> described below onthe 2 front-end: 192.168.1.10 & 192.168.1.20 >>> - Restarted the front-end >>> - Validated that :/backhand/ is displaying the correct 2 servers >>> - Use a browser on 192.168.1.101 on the same segment with No local >>> browsing cache (netscape 6.2). >>> >>> Thx >>> >>> Lionel >>> >>> >>> _______________________________________________ >>> backhand-users mailing list >>> backhand-users at lists.backhand.org >>> http://lists.backhand.org/mailman/listinfo/backhand-users >> >> >> >> >> _______________________________________________ >> backhand-users mailing list >> backhand-users at lists.backhand.org >> http://lists.backhand.org/mailman/listinfo/backhand-users > > > > _______________________________________________ > backhand-users mailing list > backhand-users at lists.backhand.org > http://lists.backhand.org/mailman/listinfo/backhand-users From gmaffia at technisys.net Thu Apr 18 02:47:15 2002 From: gmaffia at technisys.net (Gabriel Maffia) Date: Thu, 18 Apr 2002 03:47:15 -0300 Subject: [m_b_users] mod_backhand and Tomcat Message-ID: <001d01c1e6a4$e2d48c90$0202000a@homer> Hello Again: I've been testing mod_backhand with the cgi-test pages and it seemed to work fine. Howerver, when I want to configure it for my VirtualHost, it just doesn't seem to work. I am running a tomcat hosting... The Directory that I want to balance is /var/tomcat4/domains/domainname/webapps/ Is it OK if I add ... .. .. Inside the virtualhost for that app?.... I also tried making two different index.html files in /usr/local/apache/htdocs, and set on my default host BackHand byRandom BackHand byLoad but that didn't work either. Both of the machines I am using appear in the stats window but they keep in 0[0] requests served (I don't know if it is supposed to work that way). Thanks a lot to you all for all the help so far! Gabriel Maffia. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.omniti.com/pipermail/backhand-users/attachments/20020418/9e54ec88/attachment.html From alvar at a-blast.org Thu Apr 18 06:48:14 2002 From: alvar at a-blast.org (Alvar Freude) Date: Thu, 18 Apr 2002 12:48:14 +0200 Subject: [m_b_users] Virtual Hosts In-Reply-To: <1608120000.1019092576@gnarzelwicht.delirium-arts.de> References: <1608120000.1019092576@gnarzelwicht.delirium-arts.de> Message-ID: <1870960000.1019126894@gnarzelwicht.delirium-arts.de> Hi, has someone an example configuration for a cluster with x servers and name based virtual hosts? Ciao Alvar -- // STOP CENSORSHIP! http://www.odem.org/informationsfreiheit/en/ // Blaster: http://www.a-blast.org/ From sean at chittenden.org Thu Apr 18 14:57:48 2002 From: sean at chittenden.org (Sean Chittenden) Date: Thu, 18 Apr 2002 11:57:48 -0700 Subject: [m_b_users] Virtual Hosts In-Reply-To: <1870960000.1019126894@gnarzelwicht.delirium-arts.de>; from "alvar@a-blast.org" on Thu, Apr 18, 2002 at = 12:48:14PM References: <1608120000.1019092576@gnarzelwicht.delirium-arts.de> <1870960000.1019126894@gnarzelwicht.delirium-arts.de> Message-ID: <20020418115748.A38708@ninja1.internal> > has someone an example configuration for a cluster with x servers > and name based virtual hosts? Better question: does anyone have a patch for the mod_backhand source that allows mod_backhand to work with name based virtual hosts? :~) Currently it does not support this. How would you handle broadcasting out possibly hundreds of virtualhost names every second? There are problems with this scenario that aren't trivial. -sc -- Sean Chittenden From jesus at omniti.com Thu Apr 18 16:49:46 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Thu, 18 Apr 2002 16:49:46 -0400 Subject: [m_b_users] Virtual Hosts References: <1608120000.1019092576@gnarzelwicht.delirium-arts.de> <1870960000.1019126894@gnarzelwicht.delirium-arts.de> <20020418115748.A38708@ninja1.internal> Message-ID: <3CBF316A.5000906@omniti.com> Sean Chittenden wrote: >>has someone an example configuration for a cluster with x servers >>and name based virtual hosts? >> > >Better question: does anyone have a patch for the mod_backhand source >that allows mod_backhand to work with name based virtual hosts? > >:~) Currently it does not support this. How would you handle >broadcasting out possibly hundreds of virtualhost names every second? >There are problems with this scenario that aren't trivial. -sc > I have mod_backhand running on a small set of boxes with 30 or so virtualhosts. It works fine. The broadcasts are per machine, not per host. So, they _should_ only broadcast once per Apache instance no matter the number of virtual hosts you have. Backhand directives were in File, Location and Directory clauses in the the Apache configuration file. I don't understand the problem. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From sean at chittenden.org Thu Apr 18 17:05:09 2002 From: sean at chittenden.org (Sean Chittenden) Date: Thu, 18 Apr 2002 14:05:09 -0700 Subject: [m_b_users] Virtual Hosts and other ugly things In-Reply-To: <1608120000.1019092576@gnarzelwicht.delirium-arts.de>; from "alvar@a-blast.org" on Thu, Apr 18, 2002 at = 03:16:16AM References: <1608120000.1019092576@gnarzelwicht.delirium-arts.de> <1870960000.1019126894@gnarzelwicht.delirium-arts.de> <20020418115748.A38708@ninja1.internal> <3CBF316A.5000906@omniti.com> <1608120000.1019092576@gnarzelwicht.delirium-arts.de> <1870960000.1019126894@gnarzelwicht.delirium-arts.de> <1608120000.1019092576@gnarzelwicht.delirium-arts.de> Message-ID: <20020418140509.C38708@ninja1.internal> On Thu, Apr 18, 2002 at 04:49:46PM -0400, Theo Schlossnagle wrote: > Date: Thu, 18 Apr 2002 16:49:46 -0400 > From: Theo Schlossnagle > Subject: Re: [m_b_users] Virtual Hosts > > >>has someone an example configuration for a cluster with x servers > >>and name based virtual hosts? > > > >Better question: does anyone have a patch for the mod_backhand > >source that allows mod_backhand to work with name based virtual > >hosts? > > > >:~) Currently it does not support this. How would you handle > >broadcasting out possibly hundreds of virtualhost names every > >second? There are problems with this scenario that aren't trivial. > >-sc > > I have mod_backhand running on a small set of boxes with 30 or so > virtualhosts. It works fine. The broadcasts are per machine, not > per host. So, they _should_ only broadcast once per Apache instance > no matter the number of virtual hosts you have. > > Backhand directives were in File, Location and Directory clauses in the > the Apache configuration file. I don't understand the problem. I was thinking of mass virtual hosting. My bad. Having looked at his two prior posts, it looked like he wanted to say server A is handling virtual hosts X, Y, and Z. Server B had I, J, and K. Server C is acting as a reverse proxy and delegating requests to server's A and B accordingly. If you've only got a handfull of machines and the virtual hosts are specified in the configuration file, you could use byHostname to delegate requests to A and B. If you add a new virtual host, you'll have to add it to server C and HUP apache: mod_backhand doesn't broadcast the virtual hosts that it serves. Sorry for the confusion: my headspace is elsewhere. -sc -- Sean Chittenden -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 210 bytes Desc: not available Url : http://lists.omniti.com/pipermail/backhand-users/attachments/20020418/1e1f43b0/attachment.bin From kerstein at wapme-systems.de Fri Apr 19 04:33:16 2002 From: kerstein at wapme-systems.de (Martin Kerstein) Date: Fri, 19 Apr 2002 10:33:16 +0200 Subject: [m_b_users] byRandom doesn't work with two servers Message-ID: <3CBFD64C.5871168A@wapme-systems.de> Hy. I have been testing mod_backhand for a while know, and I still have a similar Problem: Everytime the requests are transferred to the 2nd Server (not the frontend server) I get the error index.whtml not found on this server. The log shows the following: All funcs executed -> callisto [notice] (Type =>) = *{ backhand-redirection, backhand:/index.whtml, backhand/redirect } If I disable mod_backhand and test each server seperately, everything works fine. I have tried byRandom and byLoad, /backhand/ shows that both servers are there and ready, and I am running out of ideas .... The setup I am using is Apache/1.3.22 (Unix) mod_ssl/2.8.5 OpenSSL/0.9.6c mod_jk/1.1.0 mod_gzip/1.3.19.1a PHP/4.1.1 DAV/1.0.3 AuthMySQL/2.20 mod_backhand/1.2.1 and I am using Tomcat 4.0.2 as servlet engine. Tomcat is loadbalanced by mod_jk. I really appreciate any help, even if it is something like "Man, you are stupid, the solution is......" :-)) By the way, please reply to my Mail adress as I am not part of this list Thank you very much Martin #Daniel Bodurant wrote: #I was having a very similar problem. #Check the apache logs, if it is trying to access the file /index.html #instead of /path/to/www/index.html, than I was having the same problem. #Search the mail list archives for similar problems. # #What was happening for me was apache was segfaulting every few requests. I #never came up with a good solution/workaround, and eventually resorted to #rebuilding the machine. -- F?r weitere Fragen stehe ich Ihnen gerne telefonisch oder via E-Mail zur Verf?gung. Mit freundlichen Gr??en Martin Kerstein Unix-Administration ----------------------------------------------------------------- Wapme Systems AG M?nsterstr. 248 40470 D?sseldorf Tel.: + 49 -211- 74845 - 0 Fax: + 49 -211- 6026 - 2801 E-Mail: kerstein at wapme-systems.de Internet:http://www.wapme-systems.de ----------------------------------------------------------------- - Wapme Systems AG - Der Wert f?r mobile Internetkommunikation am NEUEN MARKT WKN: 549550 - Reutersk?rzel: WPM G.F From ts77 at dasburo.de Mon Apr 22 08:04:23 2002 From: ts77 at dasburo.de (Thomas Seifert) Date: Mon, 22 Apr 2002 14:04:23 +0200 Subject: [m_b_users] mod_backhand for apache 2.0? Message-ID: <3CC3FC47.6E4E5BA3@dasburo.de> Hi there, just wondering if mod_backhand is available for Apache 2.0 ? If it isn't, is there some development in this direction? Thanks, Thomas From john.weir at gs.com Wed Apr 24 02:50:39 2002 From: john.weir at gs.com (Weir, John) Date: Wed, 24 Apr 2002 07:50:39 +0100 Subject: [m_b_users] Newbie question: Loadbalancing Apache and Tomcat Message-ID: <65CD76474028D311B54E0090275CC2B1100FF279@gsln31e.fw.gs.com> All, I have a geographically disparate site ( LDN, NY, TKY, HK ) each running an Apache 1.3.23 instance, with 1 or 2 tomcat 4.0.2 servlet engines at each site. I use mod_jk between the apaches and tomcat for load balancing. should I use mod_backhand for load balancing instead, and can I cluster the disparate sites ? Thoughts appreciated From yairamir at cnds.jhu.edu Wed Apr 24 08:17:46 2002 From: yairamir at cnds.jhu.edu (Yair Amir) Date: Wed, 24 Apr 2002 08:17:46 -0400 Subject: [m_b_users] Newbie question: Loadbalancing Apache and Tomcat References: <65CD76474028D311B54E0090275CC2B1100FF279@gsln31e.fw.gs.com> Message-ID: <3CC6A26A.4010405@cnds.jhu.edu> Hi, mod_backhand can do a good job balancing each of the clusters. To balance globally you need to augment it with another method. You can read some ideas on how to do that in some other work done at Hopkins/CNDS here: http://www.cnds.jhu.edu/pub/papers/web_opt.pdf The following web site has in one place several things from CNDS that might be of interest for those that need balancing and replication on the wide area network. http://www.cnds.jhu.edu/rep.html Cheers, :) Yair. http://www.cs.jhu.edu/~yairamir Weir, John wrote: > All, > > I have a geographically disparate site ( LDN, NY, TKY, HK ) each running an > Apache 1.3.23 instance, with 1 or 2 tomcat 4.0.2 servlet engines at each > site. I use mod_jk between the apaches and tomcat for load balancing. should > I use mod_backhand for load balancing instead, and can I cluster the > disparate sites ? > > Thoughts appreciated > > _______________________________________________ > backhand-users mailing list > backhand-users at lists.backhand.org > http://lists.backhand.org/mailman/listinfo/backhand-users > > > From mattcauth at yahoo.com Wed Apr 24 09:51:02 2002 From: mattcauth at yahoo.com (Matt C.) Date: Wed, 24 Apr 2002 06:51:02 -0700 (PDT) Subject: [m_b_users] cpu / memory intensive children In-Reply-To: <3CC6A26A.4010405@cnds.jhu.edu> Message-ID: <20020424135102.91909.qmail@web14001.mail.yahoo.com> I've got a single point 3 machine cluster: Things work well for the most part, but I've noticed that (in 'top') some of the apache children are substantially larger than all the others after only a few reqests; most hover around 5M (normal), but others grow wildy up to 20,30, even 50M. This instance handles static content and ssl, along with being the backhand decision maker. When one of these fat children field a request, they grow quickly in memory and the cpu pegs. When the lighter (healthy?) children handle a request the performance is great and the cpu is totally fine. Software: Solaris 8, Apache 1.3.12 (old, I know), static backhand, raven ssl. Apache was compiled with a shared core. The binary shared via NFS. A typical directe (pretty straight forward) Backhand removeSelf Backhand byAge Backhand byLoad I doubt that the above could contribute to this problem, but I've included it just in case. If anyone out there has some ideas I would appreciate it greatly. Regards, Matt __________________________________________________ Do You Yahoo!? Yahoo! Games - play chess, backgammon, pool and more http://games.yahoo.com/ From Kee at ksl.co.uk Thu Apr 25 06:16:30 2002 From: Kee at ksl.co.uk (Kee Gohil) Date: Thu, 25 Apr 2002 11:16:30 +0100 Subject: [m_b_users] Servers in .co.uk domain Message-ID: <89CABAF379428E4E9E5A940A34074F835726@zootv.ksl.co.uk> Hi All I've been running backhand for about one month now. My set up had www.oddschecker.co.uk as the front tier and www2.oddschecker.com and www3.oddschecker.com in the back tier. Front tier configuration looked like this Backhand byAge Backhand byLoad Backhand HTTPRedirectToName %3S This worked fine and I was very happy with this situation. I changed the back tier machines from .com to .co.uk so now I have www2.oddschecker.co.uk and www3.oddschecker.co.uk and the following front tier configuration. Backhand byAge Backhand byLoad Backhand HTTPRedirectToName %4S Now the load balancing doesn't work. Versions of operating system and apache have not changed. All I did was to change the hostname of the machines and the ServerName directive in the httpd.conf. What am I missing? Regards Kee Gohil Oddschecker From mjy at pobox.com Thu Apr 25 07:56:36 2002 From: mjy at pobox.com (Marinos J. Yannikos) Date: Thu, 25 Apr 2002 13:56:36 +0200 Subject: [m_b_users] Servers in .co.uk domain References: <89CABAF379428E4E9E5A940A34074F835726@zootv.ksl.co.uk> Message-ID: <3CC7EEF4.4080804@pobox.com> Hi, have you tried this without any format strings? As far as I can tell, they are only useful when you want to redirect based on parts of the original request's host. In your case, if there's only one front-end host that needs to be redirected, it isn't necessary to "know" the original host name at all. Otherwise, %-1H should probably work (if you have www2 and www3 as the back tier hosts). Regards, Marinos Kee Gohil wrote: > > Hi All > > I've been running backhand for about one month now. My set up had > www.oddschecker.co.uk as the front tier and www2.oddschecker.com and > www3.oddschecker.com in the back tier. > > Front tier configuration looked like this > > Backhand byAge > Backhand byLoad > Backhand HTTPRedirectToName %3S > > This worked fine and I was very happy with this situation. > > I changed the back tier machines from .com to .co.uk so now I have > www2.oddschecker.co.uk and www3.oddschecker.co.uk and the following > front tier configuration. > > Backhand byAge > Backhand byLoad > Backhand HTTPRedirectToName %4S > > Now the load balancing doesn't work. Versions of operating system and > apache have not changed. All I did was to change the hostname of the > machines and the ServerName directive in the httpd.conf. > > What am I missing? > > Regards > > Kee Gohil > Oddschecker > > _______________________________________________ > backhand-users mailing list > backhand-users at lists.backhand.org > http://lists.backhand.org/mailman/listinfo/backhand-users From prazak at grisoft.cz Thu Apr 25 09:22:20 2002 From: prazak at grisoft.cz (Petr Prazak) Date: Thu, 25 Apr 2002 15:22:20 +0200 Subject: [m_b_users] redirecting on overload Message-ID: <100188394437.20020425152220@grisoft.cz> Hello, I have a specific question about configuration of mod_bh. On my site, I have a web server and a cluster of file servers. From the web server I redirect requests for specific files to the file servers. I'd like to setup the mod_bh of the servers to better distribute the load of the file servers (now it is done as a dns round-robin). However, I'd like to have an option to redirect request to a different site (Akamai) if all file servers are busy (or overloaded, according to candidacy functions). I think a special candidacy function would be needed. Am I right ? -- Petr Prazak GRISOFT From jesus at omniti.com Thu Apr 25 09:59:14 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Thu, 25 Apr 2002 09:59:14 -0400 Subject: [m_b_users] redirecting on overload References: <100188394437.20020425152220@grisoft.cz> Message-ID: <3CC80BB2.6030207@omniti.com> Petr Prazak wrote: >I have a specific question about configuration of mod_bh. > >On my site, I have a web server and a cluster of file servers. From >the web server I redirect requests for specific files to the file >servers. I'd like to setup the mod_bh of the servers to better >distribute the load of the file servers (now it is done as a dns >round-robin). > >However, I'd like to have an option to redirect request to a different >site (Akamai) if all file servers are busy (or overloaded, according to >candidacy functions). > >I think a special candidacy function would be needed. Am I right ? > I'll prefix this by saying thatI haven't looked at the HTTPRedirectByName code since I wrote it :-) Well. I think I would write my own candidacy function to do exactly what I wanted, but I bet you can get something working with the existing functions. If you were running your own CDN, or used ont that didn't manipulate the URL, you could use something like: Backhand HTTPRedirectToName images.%1H However, if you are using akamai, or some such beast, you need to munge the URL. I would write it all in one candidacy funciont that changes the request uri and sets the new hostname for redirect too. You can steal it mostly from the HTTPRedirectByName code. You want the hostname to simply be rewritten to something like: a160.g.akamaitech.net and the request rewritten with the same sort of %S and %H stuff as the HTTPRedirectToName supports. /images/pixel.gif would nee dto become something like: /3/160/31/9/%0H/images/pixel.gif (which %0H would expand to the full Host: header of the request). I am _sure_ that this can be done with backhand and mod_rewrite together, but my rewrite kung fu is inferior. You could solve this by aliasing /3/160/31/9/www.example.com/images/ to /images/ on your server and then blanket rewrite every image that arrives like that. Then you really only need to swap out the host and not change the URI at all. Enough ideas... I will stop rambling now. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From john.weir at gs.com Thu Apr 25 10:03:01 2002 From: john.weir at gs.com (Weir, John) Date: Thu, 25 Apr 2002 15:03:01 +0100 Subject: [m_b_users] Newbie question: Loadbalancing Apache and Tomcat Message-ID: <65CD76474028D311B54E0090275CC2B1100FF297@gsln31e.fw.gs.com> I am a bit confused:-( sorry for being dumb. With Apache and Tomcat, I use mod_jk to loadbalance jsp requests across multiple servlet engines. This works well. If I replace mod_jk, with mod_backhand, would it be better to loadbalance requests across multiple Apaches, with each Apache associated with a single tomcat servlet engine, and as Yair suggests use an alternative strategy to route requests correctly globally? Anyone got experience of this? If so, I'd really appreciate a look at a sample config file and any tips that you may have. Thanks in advance John -----Original Message----- From: Yair Amir [mailto:yairamir at cnds.jhu.edu] Sent: Wednesday, April 24, 2002 1:18 PM To: backhand-users at lists.backhand.org Subject: Re: [m_b_users] Newbie question: Loadbalancing Apache and Tomcat Hi, mod_backhand can do a good job balancing each of the clusters. To balance globally you need to augment it with another method. You can read some ideas on how to do that in some other work done at Hopkins/CNDS here: http://www.cnds.jhu.edu/pub/papers/web_opt.pdf The following web site has in one place several things from CNDS that might be of interest for those that need balancing and replication on the wide area network. http://www.cnds.jhu.edu/rep.html Cheers, :) Yair. http://www.cs.jhu.edu/~yairamir Weir, John wrote: > All, > > I have a geographically disparate site ( LDN, NY, TKY, HK ) each running an > Apache 1.3.23 instance, with 1 or 2 tomcat 4.0.2 servlet engines at each > site. I use mod_jk between the apaches and tomcat for load balancing. should > I use mod_backhand for load balancing instead, and can I cluster the > disparate sites ? > > Thoughts appreciated > > _______________________________________________ > backhand-users mailing list > backhand-users at lists.backhand.org > http://lists.backhand.org/mailman/listinfo/backhand-users > > > _______________________________________________ backhand-users mailing list backhand-users at lists.backhand.org http://lists.backhand.org/mailman/listinfo/backhand-users From jesus at omniti.com Thu Apr 25 10:56:18 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Thu, 25 Apr 2002 10:56:18 -0400 Subject: [m_b_users] Newbie question: Loadbalancing Apache and Tomcat References: <65CD76474028D311B54E0090275CC2B1100FF297@gsln31e.fw.gs.com> Message-ID: <3CC81912.30401@omniti.com> Weir, John wrote: >I am a bit confused:-( sorry for being dumb. > >With Apache and Tomcat, I use mod_jk to loadbalance jsp requests across >multiple servlet engines. This works well. If I replace mod_jk, with >mod_backhand, would it be better to loadbalance requests across multiple >Apaches, with each Apache associated with a single tomcat servlet engine, >and as Yair suggests use an alternative strategy to route requests correctly >globally? > >Anyone got experience of this? If so, I'd really appreciate a look at a >sample config file and any tips that you may have. > >Thanks in advance > Asking questions about how to best accomplish your goals is seldom dumb and often many people benefit from observing the conversation on these mailing lists. However, I didn't really get a good feeling for your goals. Do you need to load-balancing globally? Is it really likely at all that one geographic location could get so overloaded that you would want to push someone to another geographic location? If so, you will want to use some sort of resource monitoring on the clusters to make these decisions and redirect people based on that information. mod_backhand can do this. If, on the other hand, you simply want people in the americas to go to your SJC datacenter, Eurpoe to go to your LDN datacenter and the far east to go to your TKY datacenter, then you aren't looking for load balancing. You are looking for a service resolution protocol. DNS works, but wasn't designed for this. There are a few papers and even fewer implementations that can give you the result you are looking for -- but they exist. Yair referenced a paper already. UltraDNS (www.ultradns.net) can help you out with this. They have an implementation of a global DNS system that can return different IPs based on the location of the requestor. (effectively finding the closest cluster). I think you need to sit down and clearly define the business requirements of the system. If it is a large system, it stand to reason, there will be some money in it... You should probably hire a consultant to review your goals and your proposed implementation. It is good to use someone who has done this before as a sounding board. Saves time. Saves money. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From john.weir at gs.com Thu Apr 25 12:08:52 2002 From: john.weir at gs.com (Weir, John) Date: Thu, 25 Apr 2002 17:08:52 +0100 Subject: [m_b_users] Newbie question: Loadbalancing Apache and Tomcat Message-ID: <65CD76474028D311B54E0090275CC2B1100FF298@gsln31e.fw.gs.com> Theo, I was planning to use a simple http redirect from the primary site, to the nearest local site, using info in the request header. At each site I have several servers, some running tomcat, some running apache and tomcat. I have several issues I need to address: - The JSP's are fairly intensive and experience shows that I need several servers - I would like to be able to avoid redirecting to a site that has no webserver, even if the machine is available. - Authenticated User sessions should be sticky WRT to servlet engine - Users like a single address for access to the site. - The servers are 'cheap' hardware that can fail - I have servers in multiple datacentres within a MAN. The Links are high speed, but multicast access is restricted - If I lose a whole location ( e.g. LDN ) I want to fall over to another location seamlessly - Some requests make sense to be executed in specific regional centers for access to local data Today I use mod_jk for loadbalancing the work and keeping sessions together. I assign a weighting for load at startup time, and use a statistical round-robin algorithm to 'spread' the requests. Server and site failure are dealt with manually. I am trying to workout if spending time in learning how to use mod_backhand etc will offer me a good return on investment, and seek a jump start into the process through the good people on this mailing list Thanks for the encouragement John -----Original Message----- From: Theo Schlossnagle [mailto:jesus at omniti.com] Sent: Thursday, April 25, 2002 3:56 PM To: backhand-users at lists.backhand.org Subject: Re: [m_b_users] Newbie question: Loadbalancing Apache and Tomcat Weir, John wrote: >I am a bit confused:-( sorry for being dumb. > >With Apache and Tomcat, I use mod_jk to loadbalance jsp requests across >multiple servlet engines. This works well. If I replace mod_jk, with >mod_backhand, would it be better to loadbalance requests across multiple >Apaches, with each Apache associated with a single tomcat servlet engine, >and as Yair suggests use an alternative strategy to route requests correctly >globally? > >Anyone got experience of this? If so, I'd really appreciate a look at a >sample config file and any tips that you may have. > >Thanks in advance > Asking questions about how to best accomplish your goals is seldom dumb and often many people benefit from observing the conversation on these mailing lists. However, I didn't really get a good feeling for your goals. Do you need to load-balancing globally? Is it really likely at all that one geographic location could get so overloaded that you would want to push someone to another geographic location? If so, you will want to use some sort of resource monitoring on the clusters to make these decisions and redirect people based on that information. mod_backhand can do this. If, on the other hand, you simply want people in the americas to go to your SJC datacenter, Eurpoe to go to your LDN datacenter and the far east to go to your TKY datacenter, then you aren't looking for load balancing. You are looking for a service resolution protocol. DNS works, but wasn't designed for this. There are a few papers and even fewer implementations that can give you the result you are looking for -- but they exist. Yair referenced a paper already. UltraDNS (www.ultradns.net) can help you out with this. They have an implementation of a global DNS system that can return different IPs based on the location of the requestor. (effectively finding the closest cluster). I think you need to sit down and clearly define the business requirements of the system. If it is a large system, it stand to reason, there will be some money in it... You should probably hire a consultant to review your goals and your proposed implementation. It is good to use someone who has done this before as a sounding board. Saves time. Saves money. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 _______________________________________________ backhand-users mailing list backhand-users at lists.backhand.org http://lists.backhand.org/mailman/listinfo/backhand-users From ex at valis.net.pl Thu Apr 25 12:57:03 2002 From: ex at valis.net.pl (Tomasz Szymczak) Date: Thu, 25 Apr 2002 18:57:03 +0200 Subject: [m_b_users] [OT] Replacing reverse squid with mod_proxy In-Reply-To: References: Message-ID: <20020425165703.GA2551@eldritch.valis.net.pl> 25 marzec (pon) : Hans Juergen von Lengerke wrote: >We are currently using squid set up as a reverse proxy to accelerate >several heavy backends (mod_perl, etc) and to protect them from slow >client connections. > >I am looking into replacing the squid with apache+mod_proxy. Why? >Because ultimately I'd like to be able to cluster the frontend using >mod_backhand + wackamole. The primary reason for clustering isn't for >load-balancing (yet) but for failover handling. So, ideally, one machine >should be enough to serve the whole server load. I use similar configuration, but with no backhand or wackamole (yet :). Frontend: apache+mod_proxy+mod_gzip (self compiled, static, very small) Backends: apache+php (for dynamic content), mathopd (for static files/images) Frontend is PentiumIII 700Mhz with 512MB ram. I dont use any cache stuff in mod_proxy (mathopd for static files is really fast and lightweight solution), but i compress (mod_gzip) all html pages which proxy forwards. Memory usage on frontend is very low, here are some samples: 101 apaches - 77M Active 420 apaches - 155M Active 689 apaches - 210M Active 927 apaches - 265M Active 1298 apaches - 304M Active Proxy avarage (from 10 days) is 40 requests/sec, but this number is of course much bigger at high traffic hours. apache+mod_proxy+mod_gzip is really cool solution .ex. -- Tomasz Szymczak http://gratka.pl/ http://naszemiasto.pl/ _.-. jgs GCS/M d- s-:- a-- C+++$ UL++$>++++$ P++ L++>++++ E W++ N+ w O+ '( ^{_} ( PS+ PE !Y PGP+ t- X- R tv-- b++>+++ DI D>+++ G e>++++ h! r- y? `~\`-----'\ Umys? sprawia, ?e wszystko si? zmienia, ?eby upozorowa? up?yw czasu )_)---)_) From jesus at omniti.com Thu Apr 25 13:02:38 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Thu, 25 Apr 2002 13:02:38 -0400 Subject: [m_b_users] Newbie question: Loadbalancing Apache and Tomcat References: <65CD76474028D311B54E0090275CC2B1100FF298@gsln31e.fw.gs.com> Message-ID: <3CC836AE.3040908@omniti.com> Weir, John wrote: > - The JSP's are fairly intensive and experience shows that I need >several servers > Damn JSPs. I found the same thing when I tested the "Hello World" JSP, it takes java a lot of resources to write those 11 characters to someone's browser :-D > > - I would like to be able to avoid redirecting to a site that has no >webserver, even if the machine is available. > I am not sure I understand. If the site doesn't have a web server, then don't avertise it. Are you talking about failures? I think that good HA will solve all of that. mod_backhand doesn't replace a good HA device. wackamole might, but it isn't a robust -- I like my HA devices running on hardware not intel boxes -- I have enough problems with those. My experience with running HA on Intel with wackamole has been a pleasant one, but I still prefer a pair of failover ServerIrons or something like that. The advantage there is that they are very reliable, easy to get support and you can pin somebody's balls to the wall to fix them when the break. Every tool has its place. You can run everything off a set of Extreme switches if you want. However, every different implementation has its advantages. mod_backhand has some really cool features that will _never_ be in hardware load balancers and vice-versa. The point is that none of these technologies are mutually exclusive. I have mod_backhand, ServerIrons and wackamole all running in one set up. It works fine and I get to reap the advantages from each of the technologies with no cost. > > - Authenticated User sessions should be sticky WRT to servlet engine > Very very easy. > > - Users like a single address for access to the site. > Silly users. :-) > > - The servers are 'cheap' hardware that can fail > servers are cheap, but the time invested in troubleshooting problems with shotty machinery is often underestimated. I would pay 3 times as much for 3 boxes that never crash than I would for 10 boxes that can (and do) fail frequently. Somebody has to fix them :-( > > - I have servers in multiple datacentres within a MAN. The Links are >high speed, but multicast access is restricted > This isn't necesary. I would highly suggest setting up an IPSEC tunnel (or just a GRE tunnel) between all your locations. That way you can use directed broadcast over those links. > - If I lose a whole location ( e.g. LDN ) I want to fall over to >another location seamlessly > That is tough. There are several approaches to this, but they are all very involved. You also need to more clearly define "loose". There are many different types of failures and it will dictate the approach you use. For example, if you mean, all your servers die, then you can just have a 2 machine cluster of $400 boxes that issue HTTP redirects to a fail-over location. On the other hard, if you mean that you data center could simply "disappear" or there is an act of God, then you want yo have some seamlessly rectifiable routing in place -- this really has very little to do with backhand, or web servers -- it is a network engineering issue > - Some requests make sense to be executed in specific regional >centers for access to local data > This is an issue you really need to explore before you decide on an implementation. There are a few ways to do this, but they have different pros and cons and may not be flexible in the direction you ultimately want to take your architecture. > >Today I use mod_jk for loadbalancing the work and keeping sessions together. >I assign a weighting for load at startup time, and use a statistical >round-robin algorithm to 'spread' the requests. Server and site failure are >dealt with manually. > Well, I know from personal experience that using mod_backhand to balance sticky sessions across a local cluster of Apache/Tomcat servers works like a charm. I have had that running for months without any issues. And server failure, is of course, automatic. The users loose their sessions if they are only on the Tomcat instance that crashed, but the they are immediately dropped onto another web server and "stick"there. >I am trying to workout if spending time in learning how to use mod_backhand >etc will offer me a good return on investment, and seek a jump start into >the process through the good people on this mailing list > mod_backhand is worth looking at. It will fit in your architecture and serve a valuable purpose -- though I am sure some other technology could too. The course notes and presentations on www.backhand.org might be insightful. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From jesus at omniti.com Thu Apr 25 17:02:41 2002 From: jesus at omniti.com (Theo Schlossnagle) Date: Thu, 25 Apr 2002 17:02:41 -0400 Subject: [m_b_users] cpu / memory intensive children References: <20020424135102.91909.qmail@web14001.mail.yahoo.com> Message-ID: <3CC86EF1.7000704@omniti.com> Matt C. wrote: >I've got a single point 3 machine cluster: Things work well for the most part, but >I've noticed that (in 'top') some of the apache children are substantially larger >than all the others after only a few reqests; most hover around 5M (normal), but >others grow wildy up to 20,30, even 50M. This instance handles static content and >ssl, along with being the backhand decision maker. > >When one of these fat children field a request, they grow quickly in memory and the >cpu pegs. When the lighter (healthy?) children handle a request the performance is >great and the cpu is totally fine. > >Software: >Solaris 8, Apache 1.3.12 (old, I know), static backhand, raven ssl. Apache was >compiled with a shared core. The binary shared via NFS. > >A typical directe (pretty straight forward) > > > Backhand removeSelf > Backhand byAge > Backhand byLoad > > > >I doubt that the above could contribute to this problem, but I've included it just >in case. If anyone out there has some ideas I would appreciate it greatly. > I have trouble believing that backhand is the cuplrit here. It doesn't alloc much. -- Theo Schlossnagle Principal Consultant OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ Phone: +1 301 776 6376 Fax: +1 410 880 4879 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 From mattcauth at yahoo.com Thu Apr 25 22:33:39 2002 From: mattcauth at yahoo.com (Matt C.) Date: Thu, 25 Apr 2002 19:33:39 -0700 (PDT) Subject: [m_b_users] cpu / memory intensive children In-Reply-To: <3CC86EF1.7000704@omniti.com> Message-ID: <20020426023339.38942.qmail@web14005.mail.yahoo.com> > I have trouble believing that backhand is the cuplrit here. It doesn't > alloc much. Agreed. I'm totally at a loss. Trussed the thing like mad today, had one of our strongest programmers helping - we've decided that it's some sort of system wierdness with our Solaris 8 setup or a misconfiguration, although backhand was working well with my initial tests. Here is a typical scene when it starts to eat the cpu and memory: time() = 1019758885 519: brk(0x01A28A88) = 0 519: brk(0x01A2AA88) = 0 519: time() = 1019758885 519: so_socket(2, 2, 0, "", 1) = 7 519: connect(7, 0xFFBD4CD8, 16, 1) Err#146 ECONNREFUSED 519: close(7) = 0 519: brk(0x01A2AA88) = 0 519: brk(0x01A2CA88) = 0 519: brk(0x01A2CA88) = 0 519: brk(0x01A2EA88) = 0 519: write(-1, " G E T / H T T P / 1".., 281) Err#9 EBADF 519: read(-1, 0x01A1FED0, 4096) Err#9 EBADF ...followed by MANY more brk(). Not sure why, but some children are throwing ECONNREFUSED. Backside NIC looks fine, although I didn't dig in too deeply. The backhand handler displays normally. I'll RR dns for now and test a bit more from a fresh setup; if I find anything of worth (read: not an oversight on my part somewhere) I'll follow up to the list. Thanks, Matt --- Theo Schlossnagle wrote: > Matt C. wrote: > > >I've got a single point 3 machine cluster: Things work well for the most part, > but > >I've noticed that (in 'top') some of the apache children are substantially larger > >than all the others after only a few reqests; most hover around 5M (normal), but > >others grow wildy up to 20,30, even 50M. This instance handles static content and > >ssl, along with being the backhand decision maker. > > > >When one of these fat children field a request, they grow quickly in memory and > the > >cpu pegs. When the lighter (healthy?) children handle a request the performance > is > >great and the cpu is totally fine. > > > >Software: > >Solaris 8, Apache 1.3.12 (old, I know), static backhand, raven ssl. Apache was > >compiled with a shared core. The binary shared via NFS. > > > >A typical directe (pretty straight forward) > > > > > > Backhand removeSelf > > Backhand byAge > > Backhand byLoad > > > > > > > >I doubt that the above could contribute to this problem, but I've included it > just > >in case. If anyone out there has some ideas I would appreciate it greatly. > > > I have trouble believing that backhand is the cuplrit here. It doesn't > alloc much. > > -- > Theo Schlossnagle > Principal Consultant > OmniTI Computer Consulting, Inc. -- http://www.omniti.com/ > Phone: +1 301 776 6376 Fax: +1 410 880 4879 > 1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984 > 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7 > > > > > _______________________________________________ > backhand-users mailing list > backhand-users at lists.backhand.org > http://lists.backhand.org/mailman/listinfo/backhand-users __________________________________________________ Do You Yahoo!? Yahoo! Games - play chess, backgammon, pool and more http://games.yahoo.com/ From Kee at ksl.co.uk Mon Apr 29 05:02:26 2002 From: Kee at ksl.co.uk (Kee Gohil) Date: Mon, 29 Apr 2002 10:02:26 +0100 Subject: [m_b_users] Servers in .co.uk domain Message-ID: <89CABAF379428E4E9E5A940A34074F831C11@zootv.ksl.co.uk> Hi I've tried it without format strings and it doesn't work at all. I've also tried the format string "%-3S.%1H". Same result. I switch it back to .com and it all works fine. Is there any kind of debug or log that I can turn on to help identify what is going on? Thanks Kee -----Original Message----- From: Marinos J. Yannikos [mailto:mjy at pobox.com] Sent: 25 April 2002 12:57 To: backhand-users at lists.backhand.org Subject: Re: [m_b_users] Servers in .co.uk domain Hi, have you tried this without any format strings? As far as I can tell, they are only useful when you want to redirect based on parts of the original request's host. In your case, if there's only one front-end host that needs to be redirected, it isn't necessary to "know" the original host name at all. Otherwise, %-1H should probably work (if you have www2 and www3 as the back tier hosts). Regards, Marinos Kee Gohil wrote: > > Hi All > > I've been running backhand for about one month now. My set up had > www.oddschecker.co.uk as the front tier and www2.oddschecker.com and > www3.oddschecker.com in the back tier. > > Front tier configuration looked like this > > Backhand byAge > Backhand byLoad > Backhand HTTPRedirectToName %3S > > This worked fine and I was very happy with this situation. > > I changed the back tier machines from .com to .co.uk so now I have > www2.oddschecker.co.uk and www3.oddschecker.co.uk and the following > front tier configuration. > > Backhand byAge > Backhand byLoad > Backhand HTTPRedirectToName %4S > > Now the load balancing doesn't work. Versions of operating system and > apache have not changed. All I did was to change the hostname of the > machines and the ServerName directive in the httpd.conf. > > What am I missing? > > Regards > > Kee Gohil > Oddschecker > > _______________________________________________ > backhand-users mailing list > backhand-users at lists.backhand.org > http://lists.backhand.org/mailman/listinfo/backhand-users _______________________________________________ backhand-users mailing list backhand-users at lists.backhand.org http://lists.backhand.org/mailman/listinfo/backhand-users