[Dovecot] Corrupted index cache file issues (Corrupted physical size)
Hello,
I'm running Dovecot 1.1.13 on FreeBSD 7.0-RELEASE and I'm receiving a large number of errors much like the following:
dovecot: Jun 08 09:27:04 Error: IMAP(x@x.x): Corrupted index cache file /nfs/x.x/x/Maildir/dovecot.index.cache: Corrupted physical size for uid=2284: 5312 != 5217 dovecot: Jun 08 09:27:04 Error: IMAP(x@x.x): Corrupted index cache file /nfs/x.x/x/Maildir/dovecot.index.cache: Corrupted physical size for uid=2281: 6992 != 6894 dovecot: Jun 08 09:31:17 Error: IMAP(y@y.y): Corrupted index cache file /nfs/y.y/y/Maildir/dovecot.index.cache: Corrupted physical size for uid=9241: 14545 != 14453 dovecot: Jun 08 09:37:05 Error: IMAP(z@z.z): Corrupted index cache file /nfs/z.z/z/Maildir/dovecot.index.cache: Corrupted physical size for uid=22741: 12150 != 12066
These are not occurring for all users, but it occurs frequently for the addresses that do show up in the logs. The maildirs are migrated from a previous courier server. I've tried searching the archives and can't find anything related to this error. Could someone shed some light on it? Other than the error messages, I'm not aware of anything being affected, but I'd like to be sure of it
Some info about the configuration:
Maildirs are accessed via NFS, related settings are mmap_disable: yes mail_nfs_storage: yes mail_nfs_index: yes lock_method: dotlock
Regards, Brent
A little more investigation reveals the trend is that the users receiving this error access their accounts via POP3. I assume that the UID in this error is the UID of a message. What could be causing the physical size in the cache to be incorrect?
Brent Bloxam wrote:
Hello,
I'm running Dovecot 1.1.13 on FreeBSD 7.0-RELEASE and I'm receiving a large number of errors much like the following:
dovecot: Jun 08 09:27:04 Error: IMAP(x@x.x): Corrupted index cache file /nfs/x.x/x/Maildir/dovecot.index.cache: Corrupted physical size for uid=2284: 5312 != 5217 dovecot: Jun 08 09:27:04 Error: IMAP(x@x.x): Corrupted index cache file /nfs/x.x/x/Maildir/dovecot.index.cache: Corrupted physical size for uid=2281: 6992 != 6894 dovecot: Jun 08 09:31:17 Error: IMAP(y@y.y): Corrupted index cache file /nfs/y.y/y/Maildir/dovecot.index.cache: Corrupted physical size for uid=9241: 14545 != 14453 dovecot: Jun 08 09:37:05 Error: IMAP(z@z.z): Corrupted index cache file /nfs/z.z/z/Maildir/dovecot.index.cache: Corrupted physical size for uid=22741: 12150 != 12066
These are not occurring for all users, but it occurs frequently for the addresses that do show up in the logs. The maildirs are migrated from a previous courier server. I've tried searching the archives and can't find anything related to this error. Could someone shed some light on it? Other than the error messages, I'm not aware of anything being affected, but I'd like to be sure of it
Some info about the configuration:
Maildirs are accessed via NFS, related settings are mmap_disable: yes mail_nfs_storage: yes mail_nfs_index: yes lock_method: dotlock
Regards, Brent
On Jun 8, 2009, at 7:06 AM, Brent Bloxam wrote:
Hello,
I'm running Dovecot 1.1.13 on FreeBSD 7.0-RELEASE and I'm receiving
a large number of errors much like the following:dovecot: Jun 08 09:27:04 Error: IMAP(x@x.x): Corrupted index cache
file /nfs/x.x/x/Maildir/dovecot.index.cache: Corrupted physical size
for uid=2284: 5312 != 5217
How are you saving messages to the maildir? Do these happen for new
mails also or only old ones?
Timo Sirainen wrote:
On Jun 8, 2009, at 7:06 AM, Brent Bloxam wrote:
Hello,
I'm running Dovecot 1.1.13 on FreeBSD 7.0-RELEASE and I'm receiving a large number of errors much like the following:
dovecot: Jun 08 09:27:04 Error: IMAP(x@x.x): Corrupted index cache file /nfs/x.x/x/Maildir/dovecot.index.cache: Corrupted physical size for uid=2284: 5312 != 5217
How are you saving messages to the maildir? Do these happen for new mails also or only old ones?
At the moment messages are being saved to the maildir through maildrop (called from qmail), as we're migrating from an old system. I'm not sure how I could go about pinpointing whether this affects new or old emails as I'm not aware of what actually generates this error. Any suggestions?
On Jun 9, 2009, at 11:49 AM, Brent Bloxam wrote:
Timo Sirainen wrote:
On Jun 8, 2009, at 7:06 AM, Brent Bloxam wrote:
Hello,
I'm running Dovecot 1.1.13 on FreeBSD 7.0-RELEASE and I'm
receiving a large number of errors much like the following:dovecot: Jun 08 09:27:04 Error: IMAP(x@x.x): Corrupted index cache
file /nfs/x.x/x/Maildir/dovecot.index.cache: Corrupted physical
size for uid=2284: 5312 != 5217 How are you saving messages to the maildir? Do these happen for new
mails also or only old ones?At the moment messages are being saved to the maildir through
maildrop (called from qmail), as we're migrating from an old system.
I'm not sure how I could go about pinpointing whether this affects
new or old emails as I'm not aware of what actually generates this
error. Any suggestions?
You'll see the uid=2284 above. Find the filename for that UID from
dovecot-uidlist file and see how new the file is. Also does that
filename contain ,S=5312? And what is the real size of the file?
Timo Sirainen wrote:
On Jun 9, 2009, at 11:49 AM, Brent Bloxam wrote:
Timo Sirainen wrote:
On Jun 8, 2009, at 7:06 AM, Brent Bloxam wrote:
Hello,
I'm running Dovecot 1.1.13 on FreeBSD 7.0-RELEASE and I'm receiving a large number of errors much like the following:
dovecot: Jun 08 09:27:04 Error: IMAP(x@x.x): Corrupted index cache file /nfs/x.x/x/Maildir/dovecot.index.cache: Corrupted physical size for uid=2284: 5312 != 5217 How are you saving messages to the maildir? Do these happen for new mails also or only old ones?
At the moment messages are being saved to the maildir through maildrop (called from qmail), as we're migrating from an old system. I'm not sure how I could go about pinpointing whether this affects new or old emails as I'm not aware of what actually generates this error. Any suggestions?
You'll see the uid=2284 above. Find the filename for that UID from dovecot-uidlist file and see how new the file is. Also does that filename contain ,S=5312? And what is the real size of the file?
Thanks. The errors I've checked are occurring for recent files, looks like the S= field doesn't match the real size, for example:
33568 Jun 8 23:46 1235105170.42208.abc.xyz.com,S=33498:2,RS
Looking at older e-mails though I am seeing S != real size for some e-mail. Would dovecot ever be modifying the S field, or is this actually pointing to an issue with the LDA?
On Jun 9, 2009, at 12:26 PM, Brent Bloxam wrote:
Thanks. The errors I've checked are occurring for recent files,
looks like the S= field doesn't match the real size, for example:33568 Jun 8 23:46 1235105170.42208.abc.xyz.com,S=33498:2,RS
That's the problem then.
Looking at older e-mails though I am seeing S != real size for some
e-mail. Would dovecot ever be modifying the S field, or is this
actually pointing to an issue with the LDA?
Dovecot never modifies the S= field, even when it notices that it's
wrong. So it's your LDA that writes them wrong.
On Wed, 2009-06-10 at 23:11 +0400, Max Ivanov wrote:
Dovecot never modifies the S= field, even when it notices that it's wrong. So it's your LDA that writes them wrong.
Is it ok to rename all files deleting S= and W= fields
This gives new UIDs to the messages. But that's better than not being able to access them at all :)
if no keywords was used?
There's no difference with or without keywords.
participants (3)
-
Brent Bloxam
-
Max Ivanov
-
Timo Sirainen