There have been several postings here regarding problems with the threads package on RH9; RH9 supposedly has new and/or substantially upgraded thread support. The symptoms you report are similar to those observed by other RH9 users. One user reported getting OK performance by upgrading to the latest patch level. I'd planned to migrate to RH9 which is reputed to be a substantial upgrade over RH7.3 (known to be stable) and RH8 (known to be OK) but have held off due to undertainties about the threads problem and Zope. On Thu, 8 May 2003, Troy Farrell wrote:

My first inclination is to build zope from source and see if that has the same issues.

Troy

Anand Hattiangadi wrote:

...

sorry for the repost, forgot to give the earlier one a title/subject.

On Thu, 2003-05-08 at 17:15, Anand Hattiangadi wrote:

...

I have installed Zope on a Red Hat 9 server, set ownership of everything to myself, installed CMF and CMFPlone. I can start up (running as myself), and initially everything seems to be running fine. However, after clicking around a few spots on the /manage page or a test plone site everything grinds to a halt. The Z2.log shows no more requests, stupid.log (Error severity level = 0) shows no further entries besides the usual startup INFO(0) entries, the server doesn't respond any more.

The only thing to do is to stop and start (from command line, killing process), and if that doesn't work I manually delete the temp files/lock files from the /opt/zope/var, and that does the trick.

I installed basically the same deal on RH 8 on a different machine and it worked like a charm. Do you know of any incompatibilities with RH9? Do the downloaded binaries not work? Or is it some other problem entirely?

Thanks in advance for your help,

Anand

Anand Hattiangadi <anand.hattiangadi@Sun.COM> writes: I had the same problems. There are a few suggestions in the archives of this list. The one that worked for me was to add the following in my ZOPE_HOME/start script, before the call to "python z2.py": export LD_ASSUME_KERNEL=2.2.5 HTH Felix.

Having a similar problem, Zope started and froze shortly after accessing it with a browser, and the fix worked, I had to investigate "what does this really fix" and found that the fix basically works around two components which appear to be directly related to each other. The first item is this fix forces glibc to use a deprecated stack model not the floating stack feature. And second, in RedHat at least, the Native Posix Threading Library is forced to use old LinuxThreads. This seems to have been a problem in RedHat since 7 beta, I believe, and it is also being seen in other Linux distributions especially with using databases such as Oracle, DB2, etc. It is possibly causing problems on multiple processor machines, I'm not certain of that but one report indicated the problem was with the smp kernel which I am also using. I haven't been able to keep up with all the changes in recent versions of Python but has the stacks/threading libraries been rewritten? What I found on the WEB at different sites including bugzilla.RedHat.com [my comments in brackets] Some ... [applications] ... don`t work with the new floating stack feature of the i686 version of glibc ... you may force glibc to use the deprecated stack model by setting the following environment variable: LD_ASSUME_KERNEL=2.2.5 Red Hat 7.1 enabled a floating stack feature in the glibc library that breaks the IBM JDK 1.1.8 [among other things] rpm: There is a known issue with rpm-4.2-0.69 under newer 2.5.x kernels. This is related to NPTL stuff and can be avoided by forcing rpm to use old LinuxThreads instead. To do this, run rpm like this: LD_ASSUME_KERNEL=2.2.5 rpm .. However, Red Hat kernel does come with NPTL (Native Posix Threading Library) support, which works wonders for threads on Linux. Getting NPTL into a stock 2.4 kernel ... is no trivial task. We (Red Hat) should really try to resolve this before next release. ------- Additional Comment #2 From Jakub Jelinek on 2001-02-13 07:53 ------- This is a buggy JDK (it has hardcoded knowledge about thread stacks being 2MB in size aligned to that size). We're now discussing what do with that. ------- Additional Comment #3 From Jakub Jelinek on 2001-02-17 04:02 ------- [and interestingly enough (when you don't need it)] Please note that since SuSE 8.1 was released after Maple 8, it is not an officially tested and supported platform. However, we have provided the following workaround below which has successfully allowed SuSE 8.1 users to install Maple 8: Using a text editor, comment out "export LD_ASSUME_KERNEL" found in line 1331 of the LinuxInstaller.bin file located in /tmp/maple8cd/Linux/Linux as follows: #export LD_ASSUME_KERNEL Thomas Bennett 5/12/2003 2:46:43 PM, Anand Hattiangadi <anand.hattiangadi@Sun.COM> wrote:

This worked, thank you! -Anand

Felix Ulrich-Oltean wrote:

...

Anand Hattiangadi <anand.hattiangadi@Sun.COM> writes:

I had the same problems. There are a few suggestions in the archives of this list. The one that worked for me was to add the following in my ZOPE_HOME/start script, before the call to "python z2.py":

export LD_ASSUME_KERNEL=2.2.5

HTH

Felix.

_______________________________________________ Zope maillist - Zope@zope.org http://mail.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://mail.zope.org/mailman/listinfo/zope-announce http://mail.zope.org/mailman/listinfo/zope-dev )

_______________________________________________ Zope maillist - Zope@zope.org http://mail.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://mail.zope.org/mailman/listinfo/zope-announce http://mail.zope.org/mailman/listinfo/zope-dev )

----------------------------------------------------------------- Thomas McMillan Grant Bennett Appalachian State University Computer Consultant III University Library Voice: 828 262 6587 FAX: 828 262 2797 "Windows: A 32-bit GUI on top of a 16-bit wrapper around an 8-bit interpretation of a 4-bit operating system written by a 2-bit company." -David Simmons