Re: corrupted indexes rebuilding over and over
micah wrote: Now, about 50 users have a problem where their indexes are corrupted. Dovecot tries to fix them, but for some reason it is failing. The users are seeing their mails duplicated and they cannot delete mails. I can see in the logs dovecot detecting this and trying to rebuild operations during delivery and imap login, but its repeating over and over and over. For one user, I've seen more than 4k lines in the logs about this. If i try to do doveadm force-resync -u username INBOX I get this: doveadm(nuffsaid): Warning: mdbox /n/nuffsaid/mdbox/storage: Inconsistency in map index (3494,32332 != 3494,383536) I think mdbox is sensitive to index corruption. mbox/maildir indices can be rebuilt, but data loss can result if mdbox indices are corrupted: https://wiki2.dovecot.org/MailboxFormat/dbox "Grubert, Gordon" then writes: we had the same problem after a system failure. Our problem could be solved by rm /PATH/USER/mdbox/storage/dovecot.map.index* doveadm -D -v force-resync -u USER INBOX Not sure given the above URL whether this will solve your problem of make it far worse. Better back up what you have before trying anything. Joseph Tam
Re: corrupted indexes rebuilding over and over
"Grubert, Gordon" writes: > Hi, > > we had the same problem after a system failure. Our problem could be > solved by > > rm /PATH/USER/mdbox/storage/dovecot.map.index* > doveadm -D -v force-resync -u USER INBOX I thought about doing this, but mdbox documentation[0] says: Note that with dbox the Index files actually contain significant data which is held nowhere else. Index files for both single-dbox and multi-dbox contain message flags and keywords. For multi-dbox, the index file also contains the map_uids which link (via the "map index") to the actual message data. This data cannot be automatically recreated, so it is important that Index files are treated with the same care as message data files. The 'map index' for mdbox is contained in this dovecot.map.index* files and it says that it cannot be automatically recreated, so if I nuke them, will force-resync be able to re-create them? Is force-resync not 'automatic'? thanks, micah
Re: corrupted indexes rebuilding over and over
Hi, we had the same problem after a system failure. Our problem could be solved by rm /PATH/USER/mdbox/storage/dovecot.map.index* doveadm -D -v force-resync -u USER INBOX Best regards, Gordon On 05.06.2017 18:34, micah wrote: Hi, On saturday one of our dovecot machines had an OOPS, and we had to powercycle it. When it came up, it did a filesystem fsck (ext4) and fixed a couple things, but nothing in lost+found. Now, about 50 users have a problem where their indexes are corrupted. Dovecot tries to fix them, but for some reason it is failing. The users are seeing their mails duplicated and they cannot delete mails. I can see in the logs dovecot detecting this and trying to rebuild operations during delivery and imap login, but its repeating over and over and over. For one user, I've seen more than 4k lines in the logs about this. If i try to do doveadm force-resync -u username INBOX I get this: doveadm(nuffsaid): Warning: mdbox /n/nuffsaid/mdbox/storage: Inconsistency in map index (3494,32332 != 3494,383536) doveadm(nuffsaid): Warning: fscking index file /srv/dovecot_indexes/n/nuffsaid/storage/dovecot.map.index doveadm(nuffsaid): Warning: mdbox /n/nuffsaid/mdbox/storage: rebuilding indexes doveadm(nuffsaid): Warning: fscking index file /srv/dovecot_indexes/n/nuffsaid/storage/dovecot.map.index doveadm(nuffsaid): Warning: mdbox /n/nuffsaid/mdbox/storage: Inconsistency in map index (3494,32332 != 3494,384024) doveadm(nuffsaid): Warning: fscking index file /srv/dovecot_indexes/n/nuffsaid/storage/dovecot.map.index but then if i do it again, I get the same thing, over and over and over (the numbers change a little). This is dovecot 2.2.27 with mdbox format what can I do to fix this? Its driving users mad, and they keep going over quota as their space fills. thanks! micah
