[Dovecot] Corrupted index cache file
Hello
I got this error yesterday
dovecot: Feb 11 16:07:27 Error: IMAP(dumontj): FETCH for mailbox Sent UID 7139 got too little data: 11112 vs 11160 dovecot: Feb 11 16:07:27 Error: IMAP(dumontj): Corrupted index cache file /user/dumontj/.imap/Sent/dovecot.index.cache: Broken virtual size for mail UID 7139
Does remove the cache will cure the problem ?
Thanks
On Thu, 2009-02-12 at 09:31 +0100, Frank Bonnet wrote:
Hello
I got this error yesterday
dovecot: Feb 11 16:07:27 Error: IMAP(dumontj): FETCH for mailbox Sent UID 7139 got too little data: 11112 vs 11160 dovecot: Feb 11 16:07:27 Error: IMAP(dumontj): Corrupted index cache file /user/dumontj/.imap/Sent/dovecot.index.cache: Broken virtual size for mail UID 7139
This still seems to be happening sometimes with mbox. If you ever figure out a way to reproduce it I'd like to know.
Does remove the cache will cure the problem ?
It should cure the problem itself automatically. Didn't it?
On 2/12/2009 3:31 AM, Frank Bonnet wrote:
dovecot: Feb 11 16:07:27 Error: IMAP(dumontj): FETCH for mailbox Sent UID 7139 got too little data: 11112 vs 11160 dovecot: Feb 11 16:07:27 Error: IMAP(dumontj): Corrupted index cache file /user/dumontj/.imap/Sent/dovecot.index.cache: Broken virtual size for mail UID 7139
Does remove the cache will cure the problem ?
Version? There have been lots of mbox fixes in recent versions...
--
Best regards,
Charles
On Thu, Feb 12, 2009 at 5:39 PM, Charles Marcus <CMarcus@media-brokers.com> wrote:
On 2/12/2009 3:31 AM, Frank Bonnet wrote:
dovecot: Feb 11 16:07:27 Error: IMAP(dumontj): FETCH for mailbox Sent UID 7139 got too little data: 11112 vs 11160 dovecot: Feb 11 16:07:27 Error: IMAP(dumontj): Corrupted index cache file /user/dumontj/.imap/Sent/dovecot.index.cache: Broken virtual size for mail UID 7139
Does remove the cache will cure the problem ?
Version? There have been lots of mbox fixes in recent versions...
I'm still getting this error in the 1.1.x hg tree (using a post 1.1.11 snapshot taken 6 days ago), happy to see I'm not the only one.
It's really _rare_ and difficult to reproduce. Usually, it happens when a thunderbird user tells dovecot to store a mail with multipart mime attachments in a mbox folder (mostly Sent or Trash folder).
When this error appears, the last mail attachment is actually truncated.
Filesystem here is ext3 on a two node drbd active/passive cluster (Centos 5.2)
Regards, Diego.
On Fri, 2009-02-13 at 00:10 +0100, Diego Liziero wrote:
dovecot: Feb 11 16:07:27 Error: IMAP(dumontj): FETCH for mailbox Sent UID 7139 got too little data: 11112 vs 11160 .. It's really _rare_ and difficult to reproduce. Usually, it happens when a thunderbird user tells dovecot to store a mail with multipart mime attachments in a mbox folder (mostly Sent or Trash folder).
When this error appears, the last mail attachment is actually truncated.
Any idea if the saved data simply isn't written / is truncated, instead of the following message overwriting it? i.e. has this happened to the last message in the mbox file (with uid = next_uid-1 to make sure the last message wasn't actually expunged)?
On Fri, Feb 13, 2009 at 12:32 AM, Timo Sirainen <tss@iki.fi> wrote:
On Fri, 2009-02-13 at 00:10 +0100, Diego Liziero wrote:
dovecot: Feb 11 16:07:27 Error: IMAP(dumontj): FETCH for mailbox Sent UID 7139 got too little data: 11112 vs 11160 .. When this error appears, the last mail attachment is actually truncated.
Any idea if the saved data simply isn't written / is truncated, instead of the following message overwriting it? i.e. has this happened to the last message in the mbox file (with uid = next_uid-1 to make sure the last message wasn't actually expunged)?
Sorry, no idea. There is always a delay between when the error is written in dovecot.log and when I and ask the user what happened. The truncated message is always one of the latest in the mbox, but I can't tell if it happens when the message is the last one.
Maybe a debug message that logs the latest mbox uid together with the "FETCH for mailbox Sent UID 7139 got too little data : 11112 vs 11160" message could help. Or maybe a debug extra check that reads every newly written mbox mail if it's actually written up to the last byte...
Diego.
participants (4)
-
Charles Marcus
-
Diego Liziero
-
Frank Bonnet
-
Timo Sirainen