Odd problem with cyrus.seen.NEW

[Robert Scussel]

On a similar note, we are running the same kernel with XFS and I get 
similar problems -- no solution here, but a cry for info as well.

We are running imap v.2.1.5 and sasl 2.1.7, redhat kernel 2.4.19

The problem is with the seen as well, randomly entire folders will 
suddenly be marked unread, or parts of folders, etc. In the imapd.log 
file we have:

	IOERROR: opening /var/spool/imap/users/<username>/cyrus.header: No such 
file or directory

which is another problem, as everything is located in 
/var/spool/imap/user/<username>/ (cyrus.index|cyrus.header) and our seen 
files are in /var/spool/imapdb/user/...

I don't get the other IOERRORS, at least I don't think so, as the logs 
provided were unclear.

Any help would be appreciated.

Thanks,
B


Paul Ramsey wrote:
> 
> I recently moved my imap archive to a new machine, but with an identical 
> version of cyrus imapd (1.6.24). The old machine was a linux 2.2.19 
> kernel and the new one is a 2.4.19 kernel. On both machines, the archive 
> is on an ext2 filesystem.
> 
> The new system is finding it impossible to maintain the cyrus.seen 
> files. Watching the syslog, I am seeing the following log entries 
> whenever someone does something which tries to update cyrus.seen:
> 
> Nov  4 14:28:06 hydra imapd[25117]: IOERROR: writing : No such file or 
> directory
> Nov  4 14:28:06 hydra imapd[25117]: open: user chodgson opened INBOX
> Nov  4 14:28:06 hydra imapd[25116]: IOERROR: writing : No such file or 
> directory
> Nov  4 14:28:31 hydra imapd[9857]: IOERROR: writing : Invalid argument
> 
> These errors are accompanies by the occasional dropped connections by 
> the imapd (core'ing perhaps?)
> Additionally, when I look into the user directories, I see
> 
> -rw-------    1 cyrus    mail            0 Nov  4 10:30 cyrus.seen
> -rw-------    1 cyrus    mail            0 Nov  4 14:29 cyrus.seen.NEW
> 
> Which seems to indicate that imapd is dying an unforunate death while 
> trying to do temp file -> working file process.
> 
> Yes, I am running 1.6.24. To head off any "just upgrade" comments, I 
> would like to note that all the rename() and cyrus.seen code in the 
> 2.1.x tree seems to be identical to the code in the 1.6.24 tree.
> 
> Anyone have any ideas, or suggestions for me to diagnose this?
> 
> Thanks,
> 
> Paul
> 



-- 
Robert Scussel
1024D/BAF70959/0036 B19E 86CE 181D 0912  5FCC 92D8 1EA1 BAF7 0959