[ZODB-Dev] time-stamp reduction

Thomas Guettler hv at tbz-pariv.de
Fri May 20 09:46:34 EDT 2005


Hi,

Unfortunaltly the clock was set back on one server. Now I get this during pack:

Traceback (most recent call last):
  File "/home/user/app/scripts/cron/pack.py", line 77, in ?
    main()
  File "/home/user/app/scripts/cron/pack.py", line 67, in main
    storage.pack()
  File "/home/user/app/python/lib64/python2.3/site-packages/ZEO/ClientStorage.py", line 815, in pack
    return self._server.pack(t, wait)
  File "/home/user/app/python/lib64/python2.3/site-packages/ZEO/ServerStub.py", line 79, in pack
    self.rpc.call('pack', t, wait)
  File "/home/user/app/python/lib64/python2.3/site-packages/ZEO/zrpc/connection.py", line 372, in call
    raise inst # error raised by server
ZODB.fspack.CorruptedError: 
/home/user/app/var/data.fs:299901427:time-stamp reduction: 035c9d4907706f44 <= 037cf4863546e018

Version is ZODB 3.2.1.

A search with google for "zodb timestamp reduction" reveals this thread:
http://mail.zope.org/pipermail/zodb-dev/2004-August/007826.html

Now I know that using fsrecover is a bad idea, since it drops transaction with smaller timestamps.
Is "copyTransactionsFrom()" the right thing?

Should I use the latest 3.2.x or 3.3. release to try to fix this?

There are backups. The application seems healthy since the daily unittest does not fail.

 Thomas

PS: I tried to get timestamps from the tids (035c9d4907706f44), but did not
find a suitable method in ZODB.utils. How long is the clock screw?



More information about the ZODB-Dev mailing list