[ZODB-Dev] Re: "Time travel" conflict errors in Zope

[Sidnei da Silva]

On Mon, Dec 20, 2004 at 11:54:33AM +0000, Malcolm Cleaton wrote:
| Thanks Thomas. That's interesting and useful, but I'm not sure it's the
| problem in this case.
| 
| One of my colleagues has managed to find a copy of the traceback (which
| seemed to only be displayed, and not show up in the log):
| 
| An error was encountered while publishing this resource.
| 
| *ZODB.POSException.ConflictError*
| 
| Sorry, a site error occurred.
| 
| Traceback (innermost last):
| 
|     * Module ZPublisher.Publish, line 175, in publish_module_standard
|     * Module ZPublisher.Publish, line 149, in publish
|     * Module ZPublisher.Publish, line 149, in publish
|     * Module ZPublisher.Publish, line 149, in publish
|     * Module ZPublisher.Publish, line 139, in publish
|     * Module Zope.App.startup, line 163, in zpublisher_exception_hook
|     * Module ZPublisher.Publish, line 107, in publish
|     * Module Zope.App.startup, line 222, in commit
|     * Module ZODB.Transaction, line 252, in commit
|     * Module ZODB.Connection, line 737, in tpc_vote
|     * Module ZEO.ClientStorage, line 841, in tpc_vote
|     * Module ZEO.ClientStorage, line 825, in _check_serials
| 
| ConflictError: database conflict error (oid 0x10fd16, serial this txn 
| started with 0x0359debd2d2ababb 2004-12-14 14:53:10.585995, serial 
| currently committed 0x0359defb79b34a88 2004-12-14 15:55:28.523521)
| 
| 
| This traceback was seen late in the afternoon - after 16:00 - and the
| server clock was correct (enough) at the time. So, it looks like the
| timestamp on the committed transaction is reasonable, and the mystery is
| why the "serial this txn started with" is many hours in the past.
| 
| Although I've never come across the errors when turning a clock back, I'd
| expect them to be in the opposite direction - a committed transaction
| which thinks it was in the future, rather than an open transaction which
| thinks it was opened ages ago.

For the records, I got a very similar error recently, while developing
an app. I'm not sure how I got it in this state, but for sure I didn't
change the clock. What I do remember is that I had a debug console
open (./bin/zopectl debug) and made some change and committed in
there, then did some stuff in Zope, then when trying to restart Zope I
would get the exception at *startup*, *preventing* Zope from
starting. Removing the zeo cache made the error go away. There may be
some other problem or it's just because somehow a very long
transaction can leave the ZODB in such state. In my case, the debug
console was left open for more than one day, and the dates in the
traceback were roughly apart by one day, but I didn't pay enough
attention.

-- 
Sidnei da Silva <sidnei at awkly.org>
http://awkly.org - dreamcatching :: making your dreams come true
http://www.enfoldsystems.com
http://plone.org/about/team#dreamcatcher

<glyph> spiv: FIX FTP
<spiv> glyph: I'm a little pissed atm... don't encourage me to write code :)
<sjj> spiv: pissed as in angry or pissed as in australian?