[Dovecot] Mdbox corruption

[Timo Sirainen]

On 15.8.2012, at 23.01, 3.listas at adminlinux.com.br wrote:

> I have a "ubuntu10.04 + ext4-filesystem + dovecot-2.0.13 +mdbox" configuration in my server. It works fine with ~50k accounts.
> Recently happened a filesystem corruption in a device of StMailLocation and I fix with fsck.ext4.
> After this crash, one account began to show access timeout and errors like this on force-resync:
> 
> # doveadm force-resync -u account INBOX
> doveadm(account): Warning: mdbox /path-to-mdbox/storage: rebuilding indexes
> doveadm(account): Error: Corrupted dbox file /path-to-mdbox/storage/m.83 (around offset=16): EOF reading msg header (got 0/30 bytes)
> doveadm(account): Warning: dbox: Copy of the broken file saved to /path-to-mdbox/storage/m.83.broken
> doveadm(account): Error: Corrupted dbox file /path-to-mdbox/storage/m.83 (around offset=16): EOF reading msg header (got 0/30 bytes)
> doveadm(account): Error: mdbox rebuild: Failed to fix file /path-to-mdbox/storage/m.83
> doveadm(account): Error: Corrupted dbox file /path-to-mdbox/storage/m.92 (around offset=7): Invalid dbox version
> ...
> 
> Should I worry about the m.X.broken files ?
> How I fix these broken mdboxes ?

Dovecot fixed the mdbox files automatically and wrote m.*.broken files that contain the originals, just in case the fixing didn't do something correctly so you could manually look at them.

Now, if doveadm force-resync keeps finding errors all the time it's a bug. I'd first try the latest v2.1 though, may be fixed already.