On Wed, Nov 12, 2008 at 12:21 PM, Rainer Jung <[EMAIL PROTECTED]> wrote: > Jakob Ericsson schrieb: >> >> >> -- >> Jakob Ericsson >> +46 704 533 627 >> >> 11 nov 2008 kl. 22.37 skrev Christopher Schultz >> <[EMAIL PROTECTED]>: >> >> Jakob, >> >> Jakob Ericsson wrote: >>>>> We are also experiencing this problem. >>>>> Our setup is running Windows 2003 Server with Apache 2.0.59 (no >>>>> prefork), >>>>> mod_jk 1.2.22 and Apache Tomcat/6.0.13. >>>>> >>>>> Will upgrading to latest mod_proxy_ajp in Apache httpd 2.2.X solve this >>>>> problem? >> >> If you're willing to donate some time to this, please stick with mod_jk >> and work with Rainer/Mladen to fix whatever might be wrong. Upgrading to >> the latest mod_jk is a definite requirement before you continue testing. >> >>> We are in the process of updating our system to the latest mod_jk. I >>> will give an update if this solves our problem. >>> The underlying problem is hopefully the missed multi thread flag in >>> mod_jk compile. >>> Does anyone know which issue this is in bugzilla? > > On Woindows I would not expext that to be the problem. mod_jk tries to > determine automatically during compile time, whether a multi-threaded > environment gets used and then enables thread safe mutexes. > > On some more exotic platforms like AIX this determination was broken for > some time, so some versions ago we decided to compile thread-safe by > default and add a new flag to configure to allow compiling without > thread support if you give the flag explicitely. > > On Windows it should have been always thread safe. >
We have updated a couple of our machines in the production environment to 1.2.27 and it looks quite good. Tomorrow, we will upgrade all machines and hopefully the problem will disappear. I´ll keep you posted. > Nevertheless I appreciate you update first. In case you can reproduce > the behaviour, it would be extremely helpful to have a JK log file with > debug log level. Unfortunately that is a problem for production because > of the high log volume. So if you can reproduce it easily, or when only > reproducibale under load othen on a test system, a debug level JK log > would be extremely helpful. You can make that available also only privately. > We have tried to replicate the problem in a test environment but all attempts have been unsuccessful. As you probably understand, we can not enable debug log in the production environment. The thing we see is basically the same as people have said in this thread before. First request's response is served to the second request's response. :-) And it only happens when it is one the same thread in tomcat. > I took your mail as a reminder to ask Tim Redding again for his log but > did not yet get any response. > > 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] > > -- Jakob Ericsson, JAKERI AB Tel. +46 704 533 627
