[Zope-DB] Corrupt Data.fs

[Alastair Burt]

I posted two articles to the zope@zope.org mailing list[1] about this
topic. Since I got only one oblique response, I am posting here to try to
regain my faith in the ZODB. These are the problems that have seriously
weakened my faith in the ZODB:

  1. My Data.fs got corrupted and Zope would not restart. I have no idea
     how this happend. Possible contributory causes might be no disk space
     left or two Zope instances trying to serve the same Data.fs.

  2. The fsrecover utility could not help me very much (nor
     tranalyzer). They both bomb out. Tranalyzer tells me that there is a "
     transaction lengths mismatch". Frustratingly, I can see actions in the
     Data.fs (mostly changes to a Wiki) that tranalyzer fails to recognise
     even when I give it a "-s" flag to start searching for transactions
     after the problem with the transaction length.

  3. One of the more worrying aspects of this corruption is that backups do
     not seem to be much help to me. As far as I can tell, Zope was happily
     serving requests for weeks from a database that it would not be able
     to restart from. The backups made during this time are pretty much
     unusable. I guess one way to counter this would be to kill and restart
     Zope at regular intervals.

Given these problems, can I trust the ZODB with valuable data again?

--- Alastair
 
Footnotes: 
[1]  http://lists.zope.org/pipermail/zope/2001-November/103196.html
http://lists.zope.org/pipermail/zope/2001-November/103206.html