Re: Ajp13 wrong Response
I found this post while researching the mail list archives for a problem I am having. The description Henri provides sounds exactly like what is happening with my application. I have included the original threads below. Prior to reading this posting I had not come to Henri's conclusion: Under certain conditions some request get the response for some other request I was convinced my problem was related to how I was handling sessions, or how Tomcat recycles them. But I have exhausted all resources I can think of or find. This is the only thing I have found that comes close to my problem. I will admit upfront that I am no expert in Tomcat configuration. Four or five users have contacted me with the same issue, which is basically that when they login to my app they see data intended for another user. Load is probably 40 to 50 concurrent users. I cannot reproduce the problem and it seems to be very isolated, with only a few occurences over the past week or so. My environment is Red Hat Linux 7.2, Tomcat 3.3.1, Apache 1.3.22, Postgres 7.1.3. User authentication is performed against Postgres and userid is stored in a session attribute, along with a few other attributes. In the 3.3.1 release notes bug 6234 was listed as resolved. I read the contents of the bug and tried the TestServlet you provided to reproduce the error. I still couldn't reproduce the problem which would make since if I am running a version of Tomcat that is supposed to include the fix. I also ran the TestServlet on 3.2.3 and couldn't reproduce. Although nothing indicates that the bug existed prior to 3.3. In the TestServlet you indicate that the workaround is to use response.getOutputStream() instead of response.getWriter(). I am using getOutputStream(). I realize that I may not have the same problem but would appreciate any insight you could provide. ---Original Threads- Actually, you don't need to load the system (that's the good news). It's not a synchronization problem. It looks like it is a problem with error handling, but I haven't tracked down where so I can't send a patch yet. It seems that when you get an error, the writer gets stuck in it's last successful request and always sends that one. I've reverted HttpServletResponseFacade to always throwing away it's writer on recycle (so that my developers can actually get work done) as a stop-gap and that seems to have made it go away. Of course, this is far from ideal. What you need to start seeing this is something like: public class BadServlet extends HttpServlet { protected void doGet(HttpServletRequest request, HttpServletResponse response) throws ServletException, IOException { PrintWriter out = response.getWriter(); out.println(Hello World); throw new IOException(Hi Mom!); } } - Original Message - From: GOMEZ Henri [EMAIL PROTECTED] To: [EMAIL PROTECTED] Sent: Tuesday, July 03, 2001 7:45 AM Subject: RE: Ajp13 wrong Response Very difficult to reproduce a situation under heavy load. I stressed my Apache/TC with ab during nigths without any failures but It was against simple servlets - Henri Gomez ___[_] EMAIL : [EMAIL PROTECTED](. .) PGP KEY : 697ECEDD...oOOo..(_)..oOOo... PGP Fingerprint : 9DF8 1EA8 ED53 2F39 DC9B 904A 364F 80E6 -Original Message- From: Angel Aray [mailto:[EMAIL PROTECTED]] Sent: Tuesday, July 03, 2001 2:46 PM To: [EMAIL PROTECTED] Subject: RE: Ajp13 wrong Response I'll try that, but the problem I am seeing is with tomcat as late as TC 3.3-M3. -Original Message- From: GOMEZ Henri [mailto:[EMAIL PROTECTED]] Sent: Martes, 03 de Julio de 2001 04:10 a.m. To: [EMAIL PROTECTED] Subject: RE: Ajp13 wrong Response I am having a problem which is very difficult to reproduce. Under certain conditions some request get the response for some other request. It looks like at some point probably due to some synchronization issues tomcat mixes de responses from different clients. It was a known problem in early age of TC 3.3, with RequestDispatcher. Could you try a fresh install of TC 3.3-M4 ? I hadn't seen this behavior before, but since I moved from using just tomcat to using apache + mod_jk + tomcat it has occurred a couple of times. Has anyone seen this kind of behavior before? I am using apache_1.3.20, tomcat33-m3. -- To unsubscribe, e-mail: mailto:[EMAIL PROTECTED] For additional commands, e-mail: mailto:[EMAIL PROTECTED]
Re: Ajp13 wrong Response
Vince Clark wrote: I found this post while researching the mail list archives for a problem I am having. The description Henri provides sounds exactly like what is happening with my application. I have included the original threads below. Prior to reading this posting I had not come to Henri's conclusion: Under certain conditions some request get the response for some other request Hum, it came from prehistory ;-) With ajp13 and Tomcat 3.2.x, and should really be fixed with Tomcat 3.3.1. i didn't see such behaviour for months -- To unsubscribe, e-mail: mailto:[EMAIL PROTECTED] For additional commands, e-mail: mailto:[EMAIL PROTECTED]
Re: Ajp13 wrong Response
I have been turning over every stone I can find. Still no progress. I haven't ruled out a session problem but can't come up with a good way to reproduce the error. It would have to involve some way of accepting multiple cookies and making requests using each one. Do you know of any good documentation on how TC recycles sessions? I read a posting on the comp.lang.java newsgroup that eluded to the possibility of session attributes persisting in sessions that time out and are recycled. I also haven't ruled out a problem with my code, and it would make sense that this is where the problem lies since nobody else seems to be having any similar problems. Henri Gomez wrote: Vince Clark wrote: I found this post while researching the mail list archives for a problem I am having. The description Henri provides sounds exactly like what is happening with my application. I have included the original threads below. Prior to reading this posting I had not come to Henri's conclusion: Under certain conditions some request get the response for some other request Hum, it came from prehistory ;-) With ajp13 and Tomcat 3.2.x, and should really be fixed with Tomcat 3.3.1. i didn't see such behaviour for months -- To unsubscribe, e-mail: mailto:[EMAIL PROTECTED] For additional commands, e-mail: mailto:[EMAIL PROTECTED] -- Vince Clark Global Era, LLC Software, Database, and Web Solutions (303)433-4221 office (303)455-2409 fax [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] www.globalera.net -- To unsubscribe, e-mail: mailto:[EMAIL PROTECTED] For additional commands, e-mail: mailto:[EMAIL PROTECTED]
RE: [PATCH] Ajp13 wrong Response
Henri, Check what I just checked in to TC3.3. I'm pretty sure it is a more comprehensive fix for the same issue. We had seen this internally and needed a fix for it and so I just committed my fix. Mike Anderson [EMAIL PROTECTED] 07/10/01 01:58AM I'll study carefully this one to see if he didn't broke the recovery stuff added to handle case where tomcat is restarted... Thanks - Henri Gomez ___[_] EMAIL : [EMAIL PROTECTED](. .) PGP KEY : 697ECEDD...oOOo..(_)..oOOo... PGP Fingerprint : 9DF8 1EA8 ED53 2F39 DC9B 904A 364F 80E6 -Original Message- From: William Barker [mailto:[EMAIL PROTECTED]] Sent: Monday, July 09, 2001 8:58 PM To: [EMAIL PROTECTED] Subject: [PATCH] Ajp13 wrong Response This fixes the problem reported by Angel Aray in thread 23795. The diff is against TC3.3 M4. What was happening is that the user hitting the stop button in the browser was invoking the re-try broken connection logic. This totally freaks out the proxy server who now thinks that the re-sent page is actually the page for a different request.
RE: [PATCH] Ajp13 wrong Response
I'll study carefully this one to see if he didn't broke the recovery stuff added to handle case where tomcat is restarted... Thanks - Henri Gomez ___[_] EMAIL : [EMAIL PROTECTED](. .) PGP KEY : 697ECEDD...oOOo..(_)..oOOo... PGP Fingerprint : 9DF8 1EA8 ED53 2F39 DC9B 904A 364F 80E6 -Original Message- From: William Barker [mailto:[EMAIL PROTECTED]] Sent: Monday, July 09, 2001 8:58 PM To: [EMAIL PROTECTED] Subject: [PATCH] Ajp13 wrong Response This fixes the problem reported by Angel Aray in thread 23795. The diff is against TC3.3 M4. What was happening is that the user hitting the stop button in the browser was invoking the re-try broken connection logic. This totally freaks out the proxy server who now thinks that the re-sent page is actually the page for a different request.
[PATCH] Ajp13 wrong Response
This fixes the problem reported by Angel Aray in thread 23795. The diff is against TC3.3 M4. What was happening is that the user hitting the stop button in the browser was invoking the re-try broken connection logic. This totally freaks out the proxy server who now thinks that the re-sent page is actually the page for a different request. jk_ajp13_worker.diff
RE: Ajp13 wrong Response
I am having a problem which is very difficult to reproduce. Under certain conditions some request get the response for some other request. It looks like at some point probably due to some synchronization issues tomcat mixes de responses from different clients. It was a known problem in early age of TC 3.3, with RequestDispatcher. Could you try a fresh install of TC 3.3-M4 ? I hadn't seen this behavior before, but since I moved from using just tomcat to using apache + mod_jk + tomcat it has occurred a couple of times. Has anyone seen this kind of behavior before? I am using apache_1.3.20, tomcat33-m3.
RE: Ajp13 wrong Response
I am seeing the problem appears almost on a daily basis with very little traffic. When the 3 people who use the system (consisting of 3 html forms and 3 servlets) happen to hit submit roughtly at the same time the problems appears. Once the problems appears we start to get mixed responses very frequently and the only solutions is to restart tomcat to get it back to a good state. Another, questions how would you know that ab is getting the right response. Let say your simple servlet just echo some information extracted from the httpRequest. Using ab everything would seem to be working, all request would get back some apparently good response. -Original Message- From: GOMEZ Henri [mailto:[EMAIL PROTECTED]] Sent: Martes, 03 de Julio de 2001 10:45 a.m. To: [EMAIL PROTECTED] Subject: RE: Ajp13 wrong Response Very difficult to reproduce a situation under heavy load. I stressed my Apache/TC with ab during nigths without any failures but It was against simple servlets - Henri Gomez ___[_] EMAIL : [EMAIL PROTECTED](. .) PGP KEY : 697ECEDD...oOOo..(_)..oOOo... PGP Fingerprint : 9DF8 1EA8 ED53 2F39 DC9B 904A 364F 80E6 -Original Message- From: Angel Aray [mailto:[EMAIL PROTECTED]] Sent: Tuesday, July 03, 2001 2:46 PM To: [EMAIL PROTECTED] Subject: RE: Ajp13 wrong Response I'll try that, but the problem I am seeing is with tomcat as late as TC 3.3-M3. -Original Message- From: GOMEZ Henri [mailto:[EMAIL PROTECTED]] Sent: Martes, 03 de Julio de 2001 04:10 a.m. To: [EMAIL PROTECTED] Subject: RE: Ajp13 wrong Response I am having a problem which is very difficult to reproduce. Under certain conditions some request get the response for some other request. It looks like at some point probably due to some synchronization issues tomcat mixes de responses from different clients. It was a known problem in early age of TC 3.3, with RequestDispatcher. Could you try a fresh install of TC 3.3-M4 ? I hadn't seen this behavior before, but since I moved from using just tomcat to using apache + mod_jk + tomcat it has occurred a couple of times. Has anyone seen this kind of behavior before? I am using apache_1.3.20, tomcat33-m3.
Re: Ajp13 wrong Response
I take back my previous post. This is still happening in M4 (and I'm still clueless why). - Original Message - From: Angel Aray [EMAIL PROTECTED] To: 'Tomcat-Dev [EMAIL PROTECTED] Sent: Friday, June 29, 2001 4:29 PM Subject: Ajp13 wrong Response I am having a problem which is very difficult to reproduce. Under certain conditions some request get the response for some other request. It looks like at some point probably due to some synchronization issues tomcat mixes de responses from different clients. I hadn't seen this behavior before, but since I moved from using just tomcat to using apache + mod_jk + tomcat it has occurred a couple of times. Has anyone seen this kind of behavior before? I am using apache_1.3.20, tomcat33-m3.
Re: Re: Ajp13 wrong Response
Regarding email from Bill Barker on Jun 29 18:35:
Ajp13 wrong Response
I am having a problem which is very difficult to reproduce. Under certain conditions some request get the response for some other request. It looks like at some point probably due to some synchronization issues tomcat mixes de responses from different clients. I hadn't seen this behavior before, but since I moved from using just tomcat to using apache + mod_jk + tomcat it has occurred a couple of times. Has anyone seen this kind of behavior before? I am using apache_1.3.20, tomcat33-m3.
Re: Ajp13 wrong Response
I've seen it, but couldn't track it down. I haven't seen it come back in M4, but haven't been testing M4 long either. - Original Message - From: Angel Aray [EMAIL PROTECTED] To: 'Tomcat-Dev [EMAIL PROTECTED] Sent: Friday, June 29, 2001 4:29 PM Subject: Ajp13 wrong Response I am having a problem which is very difficult to reproduce. Under certain conditions some request get the response for some other request. It looks like at some point probably due to some synchronization issues tomcat mixes de responses from different clients. I hadn't seen this behavior before, but since I moved from using just tomcat to using apache + mod_jk + tomcat it has occurred a couple of times. Has anyone seen this kind of behavior before? I am using apache_1.3.20, tomcat33-m3.