Zope 2.9.4 (but also happens on Zope.2.9.1) XP and Windows 2003 Often I see messages like the following in the event log: 2006-11-20T11:57:33 INFO ZPublisher.Conflict ConflictError at /<***some location***> : database conflict error (oid 0x03a7, class BTrees._OOBTree.OOBTree, serial this txn started with 0x0369ab934257a788 2006-11-20 16:51:15.549000, serial currently committed 0x0369ab9975fb5b99 2006-11-20 16:57:27.652000) (122 conflicts (2 unresolved) since startup at Fri Nov 17 16:01:43 2006) I am also seeing the server go into some sort hard crash (actually it doesn't crash just seems to peg the cpu doing something) ...that seems related. (but I am not totally sure). Sometimes it "magically" recovers but more often than not I have to restart the service to get some sense of normality. What does this message mean -- should I be worried about it? Is it related to my "crash"?
Often I see messages like the following in the event log:
2006-11-20T11:57:33 INFO ZPublisher.Conflict ConflictError at /<***some location***> : database conflict error (oid 0x03a7, class BTrees._OOBTree.OOBTree, serial this txn started with 0x0369ab934257a788 2006-11-20 16:51:15.549000, serial currently committed 0x0369ab9975fb5b99 2006-11-20 16:57:27.652000) (122 conflicts (2 unresolved) since startup at Fri Nov 17 16:01:43 2006) This specific message is just INFO, so it isn't visible to user, but there is a statement abotut "(2 unresolved)" conflicts, and this is worse because somebody has seen message about ConflicError.
Errors with OOBTree may be connected with session machinery. We had such problems under high load and ZEO. Changing session_resolution_seconds value in zope.conf to 300 solved the problem. You may try this but, if I remember correctly, it will cause that your sessions timeout may be 300 secs longer and/or shorter. You may try other values. It may be interesting what is <***some location***> too.
I am also seeing the server go into some sort hard crash (actually it doesn't crash just seems to peg the cpu doing something) ...that seems related. (but I am not totally sure). Sometimes it "magically" recovers but more often than not I have to restart the service to get some sense of normality. So what is a 'crash' for you? Any errors in event log or Zope just doesn't answer for a long time? If the latter thing then there is a DeadlockDebugger that shows what is Zope doing, even when it is not responsible, but I'm afraid that it won't work under Windows, but I've never tried.
What does this message mean -- should I be worried about it? As far as you see only INFO it is nothing bad, but ERROR that means conflict error was unresolved is visible to the user, although hitting 'Refresh' usually is enough.
There you may find some informations (but this is based on zope 2.7): http://www.plope.com/Members/dunny/conflicts/view -- Maciej Wisniowski
On 11/20/06, Maciej Wisniowski <maciej.wisniowski@coig.katowice.pl> wrote:
Often I see messages like the following in the event log:
2006-11-20T11:57:33 INFO ZPublisher.Conflict ConflictError at /<***some location***> : database conflict error (oid 0x03a7, class BTrees._OOBTree.OOBTree, serial this txn started with 0x0369ab934257a788 2006-11-20 16:51:15.549000, serial currently committed 0x0369ab9975fb5b99 2006-11-20 16:57:27.652000) (122 conflicts (2 unresolved) since startup at Fri Nov 17 16:01:43 2006) This specific message is just INFO, so it isn't visible to user, but there is a statement abotut "(2 unresolved)" conflicts, and this is worse because somebody has seen message about ConflicError.
Errors with OOBTree may be connected with session machinery. We had such problems under high load and ZEO. Changing session_resolution_seconds value in zope.conf to 300 solved the problem. You may try this but, if I remember correctly, it will cause that your sessions timeout may be 300 secs longer and/or shorter. You may try other values. It may be interesting what is <***some location***> too.
It seems always to be the same location (but that is the page that is always displayed for non adminstrative users so that is no suprise)
I am also seeing the server go into some sort hard crash (actually it doesn't crash just seems to peg the cpu doing something) ...that seems related. (but I am not totally sure). Sometimes it "magically" recovers but more often than not I have to restart the service to get some sense of normality. So what is a 'crash' for you?
In this case it just pegs the cpu running the python zope process at near 100%.
Any errors in event log or Zope just doesn't answer for a long time? If the latter thing then there is a DeadlockDebugger that shows what is Zope doing, even when it is not responsible, but I'm afraid that it won't work under Windows, but I've never tried.
I will have a look to see if it provides any information.
What does this message mean -- should I be worried about it? As far as you see only INFO it is nothing bad, but ERROR that means conflict error was unresolved is visible to the user, although hitting 'Refresh' usually is enough.
It seems that just having the error in info is not a problem but what I am seeing is that often after an "unresolved" instance my problem happens.
There you may find some informations (but this is based on zope 2.7): http://www.plope.com/Members/dunny/conflicts/view
I will look there to see if there is anything that seems relavent.
participants (2)
-
Brian Sullivan -
Maciej Wisniowski