Zope instanse falls asleep :-(
Hello, I have a big problem with a customers Zope instanse - almost every day it "falls asleep": When you try to access their site it's like no requests get's to the site. I try to stop the instance - and when I start it again I'm told that another process has already locked the database file. Running $ps -ax reveals that their is some stale python processes mathing the PID of the recently stopped Zope instanse ... The setup is: Debian 3.0 Zope 2.6.1 CMF 1.3 And some other CMF products. We have a similiar setup running on another machine that doesn't have these problems. Any good ideas to what is happening ??? -- Gitte Wange Technical Manager Email: gitte@mmmanager.org Web: http://www.mmmanager.org Tlf: +45 36 46 20 02 After 1.5 years of Python, I'm still discovering richness (and still unable to understand what the hell Jim Fulton is talking about). -- Gordon McMillan, 13 Mar 1998
I have a big problem with a customers Zope instanse - almost every day it "falls asleep": When you try to access their site it's like no requests get's to the site. I try to stop the instance - and when I start it again I'm told that another process has already locked the database file.
Is it spinning or not responding? There's a great HOWTO on debugging a spinning Zope on zope.org. You may have a race condition or unending recusrsion that only happens on high load. Does the event log tell you anything? Does the request log (z2.log) record requests while the Zope seems to be "asleep"?
Running $ps -ax reveals that their is some stale python processes mathing the PID of the recently stopped Zope instanse ...
This is probably Zope not responding to the kill sent by the stop script. (This suggests that it's stalled somehow.) You'll have to kill it harder. (Think kill -9 in Linux.) This will probably unlock the database file. Otherwise, you should get rid of the Data.fs.lock file. --jcc
On Thu, Oct 16, 2003 at 12:32:22PM -0500, J Cameron Cooper wrote:
Does the event log tell you anything? Does the request log (z2.log) record requests while the Zope seems to be "asleep"?
Not likely. The request log is only written to when the response is sent back to the client. I highly recommend using the -M option to z2.py. This enables debug logging from Zserver. (They grow fast, so set up proper log rotation.) This will log 4 lines for each request, telling you when the request first arrived and when it's done. -- Paul Winkler http://www.slinkp.com Look! Up in the sky! It's NEGA PERILOUS DART SCARECROW! (random hero from isometric.spaceninja.com)
Gitte Wange wrote at 2003-10-16 11:41 +0200:
I have a big problem with a customers Zope instanse - almost every day it "falls asleep": When you try to access their site it's like no requests get's to the site. I try to stop the instance - and when I start it again I'm told that another process has already locked the database file.
There is a HowTo on "SpinningZope" analysis. It may help you.
Running $ps -ax reveals that their is some stale python processes mathing the PID of the recently stopped Zope instanse ...
I have seen such behaviour in the past. It was caused by network problems. NFS access blocked deep inside the kernel and normal signals did not release the process. I would probably try to use GDB to analyse what the Zope processes are doing when they do no longer respond. Dieter
participants (4)
-
Dieter Maurer -
Gitte Wange -
J Cameron Cooper -
Paul Winkler