Our user base has grown significantly in the last few weeks. Unfortunately at hours of peak usage Zope crawls to a halt and won't even serve up pages. we have NUMBER_OF_THREADS=4 would increasing this help? Actually when we I check the processes (with ps -fe | grep py) I only see two the root thread and the thread that runs as nobody? Anybody have any suggests to help the performance of the site? Dan
I would suggest turning on debug logging (use the -M switch to z2.py) and see what methods are taking a long time (or not finishing at all)... you can use the script at http://www.zope.org/Members/mcdonc/HowTos/DEBUG-LOG to analyze the log (ignore the bit about changing z2.py at the beginning of the howto, this has been superseded by the -M switch). ----- Original Message ----- From: "Daniel Rusch" <drusch@globalcrossing.com> To: <zope@zope.org> Sent: Wednesday, April 04, 2001 2:32 PM Subject: [Zope] Help: Zope is extremely slow
Our user base has grown significantly in the last few weeks. Unfortunately at hours of peak usage Zope crawls to a halt and won't even serve up pages.
we have NUMBER_OF_THREADS=4
would increasing this help? Actually when we I check the processes (with ps -fe | grep py) I only see two the root thread and the thread that runs as nobody?
Anybody have any suggests to help the performance of the site?
Dan
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
On Wed, 4 Apr 2001, Daniel Rusch wrote:
Our user base has grown significantly in the last few weeks. Unfortunately at hours of peak usage Zope crawls to a halt and won't even serve up pages.
we have NUMBER_OF_THREADS=4
would increasing this help? Actually when we I check the processes (with ps -fe | grep py) I only see two the root thread and the thread that runs as nobody?
Anybody have any suggests to help the performance of the site?
Dan
since youre using "ps -ef" youre probably not running linux which would list threads as separate processes -- other os'es do not. i would say 4 threads is rather little, but that depends on what machine you got, how much RAM, how big your ZODB is etc. we use 20 on a fairly loaded production machine. just experiment a little -- take some spare machine, copy your application/data, load it a little (for instance with the "ab" utility that comes with apache), and try to set the number to different values. ru peter. -- _________________________________________________ peter sabaini, mailto: sabaini@niil.at -------------------------------------------------
i would say 4 threads is rather little, but that depends on what machine you got, how much RAM, how big your ZODB is etc. we use 20 on a fairly loaded production machine. just experiment a little -- take
Really mind sharing some specs on that? Any problems with so many threads? -- Andy McKay.
On Wed, 4 Apr 2001, Andy McKay wrote:
i would say 4 threads is rather little, but that depends on what machine you got, how much RAM, how big your ZODB is etc. we use 20 on a fairly loaded production machine. just experiment a little -- take
Really mind sharing some specs on that? Any problems with so many threads?
oh, not at all. is 20 many? i actually experimented with up to 80 with no problems. i'm afraid i can't give you any load numbers or througput rates (its a little while since i set it up). the machine is linux/intel pentium 3, 700mhz, 512mb, scsi raid running kernel 2.2.14. this machine belongs to zope (&apache) and some legacy data importer. our ZODB is small (~3mb) since all content is hosted in oracle, except for images which are served by apache anyway. we cache zsql queries quite extensively. currently the python process is about 30mb, i think i've seen it as big as 80mb though. the load avg. is typically around 0.1 - 0.3, ie. not much. i arrived at 20 threads simply by experimenting. throughput as measured by ab on some critical pages seemed to increase until around 20 threads, then there was no difference anymore. the site receives around 600000 pageviews / month, but we serve via reverse proxy so zope doesn't get this all. the only problem we have with the site is that it seems to loose the oracle connections from time to time, but i do not attribute that to the number of threads running since i had the thread number on a more conservative level with the same symptoms. (apart from the problem that our oracle context engine sucks a little, but thats another story :)) ru peter. -- _________________________________________________ peter sabaini, mailto: sabaini@niil.at -------------------------------------------------
the machine is linux/intel pentium 3, 700mhz, 512mb, scsi raid running kernel 2.2.14. this machine belongs to zope (&apache) and some legacy data importer. our ZODB is small (~3mb) since all content is hosted in oracle, except for images which are served by apache anyway. we cache zsql queries quite extensively. currently the python process is about 30mb, i think i've seen it as big as 80mb though. the load avg. is typically around 0.1 - 0.3, ie. not much.
Cool thats similar configuration and a similar design to our site. Im upping the number of threads as I type...
(apart from the problem that our oracle context engine sucks a little, but thats another story :))
Im having similar problems with ZODBCA, but thats another problem... :) Cheers, thanks for the advice! -- Andy McKay
participants (4)
-
Andy McKay -
Chris McDonough -
Daniel Rusch -
Peter Sabaini