|
»
"tid to old for invq" in zeo server logs?
|
View:
New views
4 Messages
—
Rating Filter:
Alert me
|
 |
"tid to old for invq" in zeo server logs?
by Chris Withers
::
Rate this Message:
Hi All,
I saw a few entries like these in the event log of a storage server that stopped responding for a couple of minutes: INFO ZEO.StorageServer (30332) tid to old for invq 252618838173181593 < 252618856850589832 ------ INFO ZEO.StorageServer (30332) tid to old for invq 252618838177204172 < 252618856850589832 What do these mean? cheers, Chris -- Simplistix - Content Management, Batch Processing & Python Consulting - http://www.simplistix.co.uk _______________________________________________ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@... https://mail.zope.org/mailman/listinfo/zodb-dev |
|
|
Re: "tid to old for invq" in zeo server logs?
by Jim Fulton
::
Rate this Message:
On Mon, Oct 19, 2009 at 6:38 AM, Chris Withers <chris@...> wrote:
> Hi All, > > I saw a few entries like these in the event log of a storage server that > stopped responding for a couple of minutes: > > INFO ZEO.StorageServer (30332) tid to old for invq 252618838173181593 < > 252618856850589832 > ------ > INFO ZEO.StorageServer (30332) tid to old for invq 252618838177204172 < > 252618856850589832 > > What do these mean? It means that the server wasn't able to provide "quick" invalidations because a client was disconnected too long. The message isn't very well worded, even aside from the misspelling. Normally, in this situation, a client would then perform a full cache verification, which puts lots of pressure on the server, possibly making it unresponsive. There's a client option "drop-cache-rather-than-verify" that can prevent this. Arguably, this is a "don't be stupid" option that should be enabled by default. Jim -- Jim Fulton _______________________________________________ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@... https://mail.zope.org/mailman/listinfo/zodb-dev |
|
|
Re: "tid to old for invq" in zeo server logs?
by Chris Withers
::
Rate this Message:
Jim Fulton wrote:
> It means that the server wasn't able to provide "quick" invalidations > because a client was disconnected too long. The message isn't very > well worded, even aside from the misspelling. Has that been fixed in the latest zodb? If not, if I get a chance, I'm happy to patch... > Normally, in this situation, a client would then perform a full cache > verification, which puts lots of pressure on the server, possibly > making it unresponsive. There's a client option > "drop-cache-rather-than-verify" that can prevent this. Was the present in Zope 2.9.8? > Arguably, this > is a "don't be stupid" option that should be enabled by default. Happy to patch this too, if it's the right thing to do and I get a chance and I'm told it's okay to do so... Chris -- Simplistix - Content Management, Batch Processing & Python Consulting - http://www.simplistix.co.uk _______________________________________________ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@... https://mail.zope.org/mailman/listinfo/zodb-dev |
|
|
Re: "tid to old for invq" in zeo server logs?
by Jim Fulton
::
Rate this Message:
On Tue, Oct 20, 2009 at 4:00 PM, Chris Withers <chris@...> wrote:
> Jim Fulton wrote: >> >> It means that the server wasn't able to provide "quick" invalidations >> because a client was disconnected too long. The message isn't very >> well worded, even aside from the misspelling. > > Has that been fixed in the latest zodb? No, the message is the same. > If not, if I get a chance, I'm happy > to patch... Sure. >> Normally, in this situation, a client would then perform a full cache >> verification, which puts lots of pressure on the server, possibly >> making it unresponsive. There's a client option >> "drop-cache-rather-than-verify" that can prevent this. > > Was the present in Zope 2.9.8? No. >> Arguably, this >> is a "don't be stupid" option that should be enabled by default. > > Happy to patch this too, if it's the right thing to do and I get a chance > and I'm told it's okay to do so... I think it's the right thing to do. Some tests would have to change too. Jim -- Jim Fulton _______________________________________________ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@... https://mail.zope.org/mailman/listinfo/zodb-dev |
| Free embeddable forum powered by Nabble | Forum Help |
