[Dovecot] Invalid old index files
Hi,
i'm in the process of evaluating dovecot. I have a rather big testing environment with four dovecot director frontend instances and four dovecot backend instances. Mailboxes are all in mdbox format and stored on nfs. Indexes are stored locally. dovecot nfs setup for high performance is used (mmap_disable = no, dotlock_use_excl = yes, mail_nfs_storage = yes, mail_nfs_index = no).
The remaining problem is, if i disable a backend server and the backend server which took over the connection already has some old index files, this index files aren't updated. If there aren't any index files all is fine and the index files are created anew. I can solve this by regularly deleting old index files via a cronjob. But i'm not sure if i made a configuration mistake and dovecot somehow can detect old index files automatically and invalidate them?
Thanks in advance Reinhard
On 5.4.2011, at 16.59, Reinhard Vicinus wrote:
i'm in the process of evaluating dovecot. I have a rather big testing environment with four dovecot director frontend instances and four dovecot backend instances. Mailboxes are all in mdbox format and stored on nfs. Indexes are stored locally.
With mdbox you must not lose indexes! They contain very important information that doesn't exist elsewhere, like what mailboxes mails exist in and what flags they have. You really need to keep index files on NFS too.
dovecot nfs setup for high performance is used (mmap_disable = no, dotlock_use_excl = yes, mail_nfs_storage = yes, mail_nfs_index = no).
With director you can keep mail_nfs_*=no. Also note that to deliver mails you need to use LMTP via directors.
participants (2)
-
Reinhard Vicinus
-
Timo Sirainen