Hi, I've got a site running a CVS version of Zope taken from the 2.6 branch on 19/9/02. The site is running off Python 2.1.3 on a Debian box and also uses ZEO 1.0. Yesterday, it appears this site died during a ZCatalog update as the last entry in Z2.log was for the advanced catalog management form. However, there was no entry for the actual re-catalog. Nothing was output to the console. Very weirdly, when I went to look at the stupid_log_file, it was totally blank. Now, I did a tail -f on that file, and when I executed the Zope stop script, the following were output to the stupid log file: ------ 2002-10-02T10:41:55 INFO(0) Z2 Caught signal SIGTERM ------ 2002-10-02T10:41:55 INFO(0) Z2 Closing all open network connections ------ 2002-10-02T10:41:55 PROBLEM(100) ClientStorage Disconnected from storage ------ 2002-10-02T10:41:56 PROBLEM(100) ClientStorage Disconnected from storage ------ 2002-10-02T10:41:56 INFO(0) client Trying to connect to server: ('x', xxxx) ------ 2002-10-02T10:41:56 PROBLEM(100) ClientStorage Disconnected from storage ------ 2002-10-02T10:41:56 INFO(0) Z2 Closing all open ZODB databases ------ 2002-10-02T10:41:56 INFO(0) ClientStorage close ------ 2002-10-02T10:41:56 PROBLEM(100) Z2 A handler for SIGTERM failed! Traceback (innermost last): Module zdaemon.SignalHandler, line 58, in signalHandler Module Signals.Signals, line 28, in shutdownHandler Module Signals.Signals, line 83, in closeall Module ZODB.DB, line 248, in close Module ZEO.ClientStorage, line 237, in close Module ZEO.zrpc, line 234, in closeIntensionally Module ZEO.trigger, line 90, in pull_trigger OSError: [Errno 32] Broken pipe When I restarted Zope, things appear to have come up fine. The funky new error log object is empty, but I guess it's not a persistent thing... The storage server (also running off the same CVS version of Zope) seems to have survived fine through all of this. Likewise, a Windows ZEO client running off the same Zope version seems to have kept going fine. Any ideas? cheers, Chris
On Wednesday 02 Oct 2002 10:50 am, Chris Withers wrote:
Yesterday, it appears this site died during a ZCatalog update as the last entry in Z2.log was for the advanced catalog management form. However, there was no entry for the actual re-catalog.
That makes sense - Z2.log entries are made once the method has completed (because it needs to include a field for the number of bytes sent back). It looks like it stalled in the middle of that method.
Very weirdly, when I went to look at the stupid_log_file, it was totally blank. Now, I did a tail -f on that file, and when I executed the Zope stop script, the following were output to the stupid log file:
2002-10-02T10:41:55 INFO(0) Z2 Closing all open network connections
Zope closed all the network connections in the asyncore map, including the one needed by ZEO that caused the traceback below. There is a fishbowl proposal to fix this, titled 'extra clean shutdown' iirc.
2002-10-02T10:41:56 PROBLEM(100) Z2 A handler for SIGTERM failed! Traceback (innermost last): Module zdaemon.SignalHandler, line 58, in signalHandler Module Signals.Signals, line 28, in shutdownHandler Module Signals.Signals, line 83, in closeall Module ZODB.DB, line 248, in close Module ZEO.ClientStorage, line 237, in close Module ZEO.zrpc, line 234, in closeIntensionally Module ZEO.trigger, line 90, in pull_trigger OSError: [Errno 32] Broken pipe
When I restarted Zope, things appear to have come up fine. The funky new error log object is empty, but I guess it's not a persistent thing...
Its not persistent. You can get it to copy errors to zLOG, which eventally targets the stupid log. Do you have that box ticked?
The storage server (also running off the same CVS version of Zope) seems to have survived fine through all of this.
hurrah.
Toby Dickenson wrote:
Yesterday, it appears this site died during a ZCatalog update as the last entry in Z2.log was for the advanced catalog management form. However, there was no entry for the actual re-catalog.
That makes sense - Z2.log entries are made once the method has completed (because it needs to include a field for the number of bytes sent back).
Yeah, that's what I thought.
It looks like it stalled in the middle of that method.
Indeed.
Very weirdly, when I went to look at the stupid_log_file, it was totally blank.
I'm still very concerned as to how this managed to happen...
2002-10-02T10:41:55 INFO(0) Z2 Closing all open network connections
Zope closed all the network connections in the asyncore map, including the one needed by ZEO that caused the traceback below. There is a fishbowl proposal to fix this, titled 'extra clean shutdown' iirc.
So, in effect, this is a bug but nothing to worry about?
When I restarted Zope, things appear to have come up fine. The funky new error log object is empty, but I guess it's not a persistent thing...
Its not persistent. You can get it to copy errors to zLOG, which eventally targets the stupid log. Do you have that box ticked?
No, I do now though... My main concern is finding out why it ground to a halt. If this is a bug in 2.6, it'd be good to get ti fixed before 2.6 final comes out... cheers, Chris
participants (2)
-
Chris Withers -
Toby Dickenson