[Mailman-Users] Bad Marshal Data

[Barry A. Warsaw]

>>>>> "JB" == Jonah Baker <jonah at votenader.org> writes:

    JB> A list of 40,000 has gotten corrupted and I can't seem to
    JB> restore it from backup.  The error I get from running check_db
    JB> is "bad marshal data."  Both config.db and config.db.last were
    JB> corrupted.  I tried replacing them with three backups but I
    JB> continue to get the "bad marshal data" error.  How do I
    JB> recover the database?  Rebuilding will be a major undertaking,
    JB> and I will loose everyone's passwords, so I would rather
    JB> massage out any errors by hand in the corrupted file.

    JB> I was told that withlist will allow me to look into a
    JB> corrupted database but I can't seem to make it work at the
    JB> command line.

    JB> What might have lead to the corruption is the rampant creation
    JB> of .tmp.  files.  There were 65 of them when I last backed up.
    JB> Any idea what might have been causing this?

I have a suspicion that disk i/o errors, such as a disk or quota
filling up could cause this.  The very latest CVS has a patch for this
which should solve the problem (which we ran into on python.org, just
as we were getting our access back!).  I've tested it with quota
exceptions on Linux.

-Barry