Need your urgent advice

[Bron Gondwana]

On Sat, Dec 8, 2012, at 10:34 AM, anant at isac.gov.in wrote:
> ----- Message from brong at fastmail.fm ---------
>      Date: Sat, 08 Dec 2012 10:18:29 +0100
>      From: Bron Gondwana <brong at fastmail.fm>
>   Subject: Re: Need your urgent advice
>        To: anant at isac.gov.in, Dan White <dwhite at olp.net>
>        Cc: Info Cyrus <info-cyrus at lists.andrew.cmu.edu>
> 
> 
> > On Sat, Dec 8, 2012, at 09:11 AM, anant at isac.gov.in wrote:
> >> Thanks for your inputs.  As per the thread link you mentioned, it
> >> looks like corruption in mailboxes.db.  I plan to start the process
> >> given in next 4 hours time with a downtime
> >> (README.HOWTO-recover-mailboxes.db).  This is going to take a huge
> >> amount of time.  I have done it before some few years back, when I had
> >> to move mailboxes to another storage.  But, this time, I did not do,
> >> since, I thought 2.3.7 and 2.3.16 are same level versions and hence
> >> may not be required.
> >>
> >> If I can get any other alternative option from you within next 4
> >> hours, then I can think of that solution.  Otherwise, I am going ahead
> >> with downtime and run the procedure documented in
> >> README.HOWTO-recover-mailboxes.db.  Any other advise from experts is
> >> welcome.
> >
> > The problem is that the mailboxes.db contents format is unchanged, but
> > if you have mailboxes.db itself as a Berkeley DB file, and your Cyrus is
> > compiled against a different version of Berkeley, then that could cause
> > problems.
> >
> > I think following the HOWTO as you plan is the best solution in your
> > circumstance.
> >
> > Regards,
> >
> > Bron.
> > --
> >   Bron Gondwana
> >   brong at fastmail.fm
> 
> Thanks.  Apart from this, should I also do
> 1. ctl_cyrusdb -r

This should be part of your START block in cyrus.conf.  You don't need
to (and shouldn't!) run it during operations.

> 2. tls_prune
> 3. ctl_cyrusdb -c
> 4. cyr_expire -E 3

These should all run from your EVENTS block in cyrus.conf (or from cron
if you're so inclined) on a scheduled basis.  I can't see that any of them
would make any difference to crashing.

> Or reconstruct -r -f will take care of all these steps?

That will get you all the mailboxes.db records correct, I hope!  I really
don't like working with berkeley DB.  Another option, if you have access to
the old copy still, is to dump the mailboxes.db from there as a plaintext
file with ctl_mboxlist and then import it into the new server.  In 2.3.16 I
would recommend skiplist for mailboxes.db.

If you're using licenced Redhat, you may also want to check with Redhat
support for advice, since they have presumably done this before.

Regards,

Bron.
-- 
  Bron Gondwana
  brong at fastmail.fm