Re: Load balancing mark down question
The release number 1.2.19 specifies a full release, not just a set of patches. So get a copy of mod_jk 1.2.19 (either binary or source and compile by yourself) and you will end up with a new mod_jk module file, you can replace in your web server. Regards, Rainer Sharma, Siddharth schrieb: Hey Rainer I currently have 1.2.15. Do I need to install all the interim releases or 1.2.19 is all encompassing? Thanks Sidd Ref: http://www.nabble.com/Load-balancing-mark-down-question-t2590766.html -Original Message- From: Rainer Jung [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 08, 2006 11:52 AM To: Tomcat Users List Subject: Re: Load balancing mark down question Sharma, from the 1.2.19 code the request should also work by completely omitting the id parameter and only using the worker name :) Regards, Rainer - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Load balancing mark down question
Sharma, from the 1.2.19 code the request should also work by completely omitting the id parameter and only using the worker name :) Regards, Rainer Sharma, Siddharth schrieb: Hi Rainer Thx for the response. What does missing answers mean? If there is no http body, will it be marked down? Yea, I figured that status is probably the best way to do it. I already have status configured and I figured out the request parameters. Only complication is the id parameter. Even if I provide the w parameter as the worker name, it does not update its status unless the 'correct' id value is provided. id value seems to be a positional index of the worker in the configuration or something. This sort of sucks for me because my coding will not be straightforward and I will have to maintain the mapping of worker name and id. -Original Message- From: Rainer Jung [mailto:[EMAIL PROTECTED] Sent: Tuesday, November 07, 2006 3:58 PM To: Tomcat Users List Subject: Re: Load balancing mark down question Hi, Sharma, Siddharth schrieb: Hi We have a cluster of Linux (RedHat) machines each housing an apache, 4 tomcat instances and 4 other jvms that run our custom servers. Each tomcat has a corresponding custom server that it delegates requests to. In other words, there is one to one correspondence between a tomcat instance and a custom server instance. Problem is that when a custom server crashes and its tomcat is still accepting, mod_jk does not mark it down. Consequently users continue to be directed to that tomcat even though they are experiencing errors. In mod_jk, we have defined a load balancing worker that fronts the 4 tomcat workers. So questions are: 1. On what basis does mod_jk mark a tomcat worker down? Is it http return codes? Never http return code. jk detects communication failures betwenn apache and tomcat: - timeouts - connection problems - missing answers - client (browser) aborts 2. Some load balancers allow scraping of the http body to see if there is an error in addition to relying on http codes. Is that possible in mod_jk? Is it configurable? No. You need to do that externally. 3. Is it possible to stop|start a mod_jk worker via an http call? Or some other programmable way? Yes. Configure a status worker: worker.list=admin worker.admin.type=status The name of the worker (here: admin) is arbitrary. Map the worker to a url in your web server, e.g. using JkMount in Apache. Add access control to the URL (using web server methods). Then open the URL in your browser and get used to its features. You can click on the individual worker members of a balancer and a form will pop up, allowing to disable (no more new sessions) or stop (no more requests) this worker. The request when sending the form is a GET request, so you can learn the URL syntax easily. Be careful: these changes are not persistant. Apache restart, even graceful, throws you back to your config file settings. Regards, Rainer Any ideas on how to solve this problem? Thanks Sidd - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: Load balancing mark down question
Hey Rainer I currently have 1.2.15. Do I need to install all the interim releases or 1.2.19 is all encompassing? Thanks Sidd Ref: http://www.nabble.com/Load-balancing-mark-down-question-t2590766.html -Original Message- From: Rainer Jung [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 08, 2006 11:52 AM To: Tomcat Users List Subject: Re: Load balancing mark down question Sharma, from the 1.2.19 code the request should also work by completely omitting the id parameter and only using the worker name :) Regards, Rainer - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Load balancing mark down question
Hi, Sharma, Siddharth schrieb: Hi We have a cluster of Linux (RedHat) machines each housing an apache, 4 tomcat instances and 4 other jvms that run our custom servers. Each tomcat has a corresponding custom server that it delegates requests to. In other words, there is one to one correspondence between a tomcat instance and a custom server instance. Problem is that when a custom server crashes and its tomcat is still accepting, mod_jk does not mark it down. Consequently users continue to be directed to that tomcat even though they are experiencing errors. In mod_jk, we have defined a load balancing worker that fronts the 4 tomcat workers. So questions are: 1. On what basis does mod_jk mark a tomcat worker down? Is it http return codes? Never http return code. jk detects communication failures betwenn apache and tomcat: - timeouts - connection problems - missing answers - client (browser) aborts 2. Some load balancers allow scraping of the http body to see if there is an error in addition to relying on http codes. Is that possible in mod_jk? Is it configurable? No. You need to do that externally. 3. Is it possible to stop|start a mod_jk worker via an http call? Or some other programmable way? Yes. Configure a status worker: worker.list=admin worker.admin.type=status The name of the worker (here: admin) is arbitrary. Map the worker to a url in your web server, e.g. using JkMount in Apache. Add access control to the URL (using web server methods). Then open the URL in your browser and get used to its features. You can click on the individual worker members of a balancer and a form will pop up, allowing to disable (no more new sessions) or stop (no more requests) this worker. The request when sending the form is a GET request, so you can learn the URL syntax easily. Be careful: these changes are not persistant. Apache restart, even graceful, throws you back to your config file settings. Regards, Rainer Any ideas on how to solve this problem? Thanks Sidd - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: Load balancing mark down question
Hi Rainer Thx for the response. What does missing answers mean? If there is no http body, will it be marked down? Yea, I figured that status is probably the best way to do it. I already have status configured and I figured out the request parameters. Only complication is the id parameter. Even if I provide the w parameter as the worker name, it does not update its status unless the 'correct' id value is provided. id value seems to be a positional index of the worker in the configuration or something. This sort of sucks for me because my coding will not be straightforward and I will have to maintain the mapping of worker name and id. -Original Message- From: Rainer Jung [mailto:[EMAIL PROTECTED] Sent: Tuesday, November 07, 2006 3:58 PM To: Tomcat Users List Subject: Re: Load balancing mark down question Hi, Sharma, Siddharth schrieb: Hi We have a cluster of Linux (RedHat) machines each housing an apache, 4 tomcat instances and 4 other jvms that run our custom servers. Each tomcat has a corresponding custom server that it delegates requests to. In other words, there is one to one correspondence between a tomcat instance and a custom server instance. Problem is that when a custom server crashes and its tomcat is still accepting, mod_jk does not mark it down. Consequently users continue to be directed to that tomcat even though they are experiencing errors. In mod_jk, we have defined a load balancing worker that fronts the 4 tomcat workers. So questions are: 1. On what basis does mod_jk mark a tomcat worker down? Is it http return codes? Never http return code. jk detects communication failures betwenn apache and tomcat: - timeouts - connection problems - missing answers - client (browser) aborts 2. Some load balancers allow scraping of the http body to see if there is an error in addition to relying on http codes. Is that possible in mod_jk? Is it configurable? No. You need to do that externally. 3. Is it possible to stop|start a mod_jk worker via an http call? Or some other programmable way? Yes. Configure a status worker: worker.list=admin worker.admin.type=status The name of the worker (here: admin) is arbitrary. Map the worker to a url in your web server, e.g. using JkMount in Apache. Add access control to the URL (using web server methods). Then open the URL in your browser and get used to its features. You can click on the individual worker members of a balancer and a form will pop up, allowing to disable (no more new sessions) or stop (no more requests) this worker. The request when sending the form is a GET request, so you can learn the URL syntax easily. Be careful: these changes are not persistant. Apache restart, even graceful, throws you back to your config file settings. Regards, Rainer Any ideas on how to solve this problem? Thanks Sidd - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Load balancing mark down question
Good point concerning the id. I'm already thinking about a couple of changes to the status worker, especially the URLs it uses and the missing parameter validation. So it's not totally unlikely, that we will have a 1.2.20 release this year inclusing better support for automating status worker actions. Missing answer: if there is no http body in the request, and the announced content length is bigger than zero, I can imagine three cases: 1) The connection is being closed by the tomcat side. This will be detected as an error. 2) You've got a reply_timeout and you run into the timeout. This will be detected as an error. 3) You've got no reply_timeout and the connection is kept alive, but no packet returned. I guess, this will just keep apache reading. I'm not sure, if the 300 seconds timeout of apache works in this case, and what will happen, after you run into this timeout. If your observations differ from my predictions, let me know :) Regards, Rainer Sharma, Siddharth schrieb: Hi Rainer Thx for the response. What does missing answers mean? If there is no http body, will it be marked down? Yea, I figured that status is probably the best way to do it. I already have status configured and I figured out the request parameters. Only complication is the id parameter. Even if I provide the w parameter as the worker name, it does not update its status unless the 'correct' id value is provided. id value seems to be a positional index of the worker in the configuration or something. This sort of sucks for me because my coding will not be straightforward and I will have to maintain the mapping of worker name and id. -Original Message- From: Rainer Jung [mailto:[EMAIL PROTECTED] Sent: Tuesday, November 07, 2006 3:58 PM To: Tomcat Users List Subject: Re: Load balancing mark down question Hi, Sharma, Siddharth schrieb: Hi We have a cluster of Linux (RedHat) machines each housing an apache, 4 tomcat instances and 4 other jvms that run our custom servers. Each tomcat has a corresponding custom server that it delegates requests to. In other words, there is one to one correspondence between a tomcat instance and a custom server instance. Problem is that when a custom server crashes and its tomcat is still accepting, mod_jk does not mark it down. Consequently users continue to be directed to that tomcat even though they are experiencing errors. In mod_jk, we have defined a load balancing worker that fronts the 4 tomcat workers. So questions are: 1. On what basis does mod_jk mark a tomcat worker down? Is it http return codes? Never http return code. jk detects communication failures betwenn apache and tomcat: - timeouts - connection problems - missing answers - client (browser) aborts 2. Some load balancers allow scraping of the http body to see if there is an error in addition to relying on http codes. Is that possible in mod_jk? Is it configurable? No. You need to do that externally. 3. Is it possible to stop|start a mod_jk worker via an http call? Or some other programmable way? Yes. Configure a status worker: worker.list=admin worker.admin.type=status The name of the worker (here: admin) is arbitrary. Map the worker to a url in your web server, e.g. using JkMount in Apache. Add access control to the URL (using web server methods). Then open the URL in your browser and get used to its features. You can click on the individual worker members of a balancer and a form will pop up, allowing to disable (no more new sessions) or stop (no more requests) this worker. The request when sending the form is a GET request, so you can learn the URL syntax easily. Be careful: these changes are not persistant. Apache restart, even graceful, throws you back to your config file settings. Regards, Rainer Any ideas on how to solve this problem? Thanks Sidd - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
