[Dovecot] blocked for xx secs
Hello,
Occasionally (a few times a day) we get batches of Warnings in the log such as: dovecot: Warning: chdir(/maildir/x.co.uk/users/y) blocked for 18 secs This tends to last a minute or so for different users.
/maildir is a NFS mount, and I'm suspecting it's the NFS server being busy and dovecot having to wait.
Other than replacing the NFS server with faster hardware is there anything else we can do?
Indexes are local, and Maildir is on the NFS.
Dovecot 1.2.5 using: mail_nfs_storage = yes mmap_disable = no
NFS mounted as; /maildir on /.automount/a.diskless/root/mail type nfs (nosuid,nodev,noatime,vers=3,proto=tcp)
TIA
Andrew
On Mon, 2010-01-25 at 13:16 +0000, Andrew Hearn wrote:
dovecot: Warning: chdir(/maildir/x.co.uk/users/y) blocked for 18 secs This tends to last a minute or so for different users.
/maildir is a NFS mount, and I'm suspecting it's the NFS server being busy and dovecot having to wait.
Right.
Other than replacing the NFS server with faster hardware is there anything else we can do?
Nope. Although with v2.0 the warning is gone, it'll just silently wait for however long it needs to. But then it's the imap/pop3 process that waits, not a forked "dovecot" master process, which was causing confusion.
Oh, and of course you could try to reduce disk I/O on the NFS server. If you've only a single Dovecot server, you can keep using mail_nfs_storage=no. Also you could try maildir_very_dirty_syncs=yes if you've nothing else besides Dovecot touching your maildirs.
participants (2)
-
Andrew Hearn
-
Timo Sirainen