RE: [Zope] Performance problems after upgrade
I don't use Mysql ... And I use Zope 2.7.2 with Python 2.3.3. But I have the RH7.3 with the same kernel ... On a 4 CPU machine. I DID tweak the interpreter lock timing to try and improve performance ... (Related to the GIL). I used the guidlines from: http://www.zope.org/Members/glpb/solaris/report_ps I have to admit I never truly benchmarked it, I just noticed it seemed to perform better. I'm not really spending too much time on it, as my focus now is to use single CPU machines for all things Zope/Python. I've never noticed any unexplained "bursty" performance, we i do have it, it's usually due to traffic, bursts of requests, and so on ... Today for instance I've been averaging 10 hits/sec and the site is nice and fast. With a 2 cpu machine you're less likely to feel the effects of the GIL issue ... In your case I don't know what I'd suspect ... You might have to figure out exactly which requests/threads run for too long and figure out what they have in common. (Which you can normally figure out just by looking at the debugging screen). If worst comes to worst, do some profiling? J.F. -----Original Message----- From: zope-bounces@zope.org [mailto:zope-bounces@zope.org]On Behalf Of Dennis Allison Sent: August 25, 2004 3:21 PM To: zope@zope.org Subject: [Zope] Performance problems after upgrade I upgraded from mysql-3.23.56-1.9 to mysql--4.0.20-0 using the mysql.org RPMS, from Zope 2.6.2b3 (Python 2.1.3) to Zope 2.6.4 (Python 2.3.4) under Linux RH7.3 with kernel 2.4.20-28.7. Host is a dual Athalon processor with 4GB of memory, 3ware RAID (configured as RAID 10) and not much else. Performance is way down from the earlier configuration with "bursty" slowdowns. There is some evidence that it is somehow related to redirection in the Zope application, but we have not been able to locate the problem. Has anyone else experienced this sort of problem? Pointers and suggestions would be appreciated. It is possible that the behavior could arise from a different kernel (and different scheduler) and the GIL (Global Interpreter Lock). We have several years experience running Zope on a dual processor machine on RH7.3 without observable GIL problems, but it is a possible source of the problem. Does anyone have a suggestion as to how to determine if that is the case? _______________________________________________ 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 )
Thanks. Your experience is encouraging. I still syspect the GIL may be one of the culprits, but I may wait a bit before pointing fingers. In the background is a move to a 2.6 kernel with cpu_affinity etc etc. I have done some profiling but the results are difficult to interpret. Andy Dustman suggested I move up to a later MySQL adaptor which I will do this evening. And the quest continues... On Wed, 25 Aug 2004 Jean-Francois.Doyon@CCRS.NRCan.gc.ca wrote:
I don't use Mysql ... And I use Zope 2.7.2 with Python 2.3.3.
But I have the RH7.3 with the same kernel ... On a 4 CPU machine.
I DID tweak the interpreter lock timing to try and improve performance ... (Related to the GIL).
I used the guidlines from:
http://www.zope.org/Members/glpb/solaris/report_ps
I have to admit I never truly benchmarked it, I just noticed it seemed to perform better.
I'm not really spending too much time on it, as my focus now is to use single CPU machines for all things Zope/Python.
I've never noticed any unexplained "bursty" performance, we i do have it, it's usually due to traffic, bursts of requests, and so on ... Today for instance I've been averaging 10 hits/sec and the site is nice and fast.
With a 2 cpu machine you're less likely to feel the effects of the GIL issue ... In your case I don't know what I'd suspect ... You might have to figure out exactly which requests/threads run for too long and figure out what they have in common. (Which you can normally figure out just by looking at the debugging screen).
If worst comes to worst, do some profiling?
J.F.
-----Original Message----- From: zope-bounces@zope.org [mailto:zope-bounces@zope.org]On Behalf Of Dennis Allison Sent: August 25, 2004 3:21 PM To: zope@zope.org Subject: [Zope] Performance problems after upgrade
I upgraded from mysql-3.23.56-1.9 to mysql--4.0.20-0 using the mysql.org RPMS, from Zope 2.6.2b3 (Python 2.1.3) to Zope 2.6.4 (Python 2.3.4) under Linux RH7.3 with kernel 2.4.20-28.7. Host is a dual Athalon processor with 4GB of memory, 3ware RAID (configured as RAID 10) and not much else.
Performance is way down from the earlier configuration with "bursty" slowdowns. There is some evidence that it is somehow related to redirection in the Zope application, but we have not been able to locate the problem.
Has anyone else experienced this sort of problem? Pointers and suggestions would be appreciated.
It is possible that the behavior could arise from a different kernel (and different scheduler) and the GIL (Global Interpreter Lock). We have several years experience running Zope on a dual processor machine on RH7.3 without observable GIL problems, but it is a possible source of the problem. Does anyone have a suggestion as to how to determine if that is the case?
_______________________________________________ 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 )
participants (2)
-
Dennis Allison -
Jean-Francois.Doyon＠CCRS.NRCan.gc.ca