Re: reoccuring DBERRORs

2006-02-24 Thread mlgw-2k5 

Simon Matter wrote:


I guess your cyrus configdirectory is /var/lib/cyrus. Then, did you try
removing the transaction logs in /var/lib/cyrus/db after removing
deliver.db and tls_cache.db?


No, I didn't. I've thought about it, but none of the messages I've read 
in the archives mentioned deleting the contents of /var/lib/cyrus/db, 
too, and I wasn't too willing to try it out and probably cause even more 
trouble.


Let's see, I'm running with Debian defaults, so the DBs should be:

- deliver.db: BDB3
- mailboxes.db: skiplist
- tls_cache.db: BDB3 (TLS caching is now disabled)
- quota: plaintext
- seen: skiplist

When I shut down Cyrus cleanly, which of the files in /var/lib/cyrus/db 
can be deleted safely along with deliver.db then?


--


Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html

reoccuring DBERRORs

2006-02-23 Thread mlgw-2k5 
Hello there,

I'm running Debian Sarge with Cyrus 2.1.18 on an IBM xSeries Dual-Xeon with
8GB RAM and 450GB disks (ServeRaid controller, RAID5 w/ hot-spare). The
machine serves 22000 accounts (mostly POP3, ca. 250 IMAP users) and has been
running happily without any notable load for a year.

When one of the RAID disks failed, the following rebuild, propably due to a
not very sensible default setting in the RAID controller, caused the machine
to stall. Authentication took forever and pop3d and imapd processes were
piling up. The server had to be shutdown rather ungracefully making the
various Cyrus DBs and some filesystems suffer.

After a successfull RAID rebuild, filesystem checks and a longish
cyrreconstruct run (7 hours) things seemed to be fine again. But since then
the system is causing trouble on a daily basis with issues like these:

cyrus/lmtpd[29054]:
  DBERROR: init /var/lib/cyrus/db: cyrusdb error
cyrus/lmtpd[29054]:
  FATAL: lmtpd: unable to init duplicate delivery database
cyrus/pop3d[29049]:
  DBERROR: init (TLS: /var/lib/cyrus/db): cyrusdb error
cyrus/lmtpd[29070]:
  DBERROR: dbenv-open '/var/lib/cyrus/db' failed:
  DB_RUNRECOVERY: Fatal error, run database recovery

Removing deliver.db and tls_cache.db between stopping and re-starting Cyrus
has worked till now, but the same thing breaking again and again really
starts to annoy me and the users.

I've run cyrreconstruct. I've wandered through the mailing list archive and
found countless posts mentioning DB errors, but no real solution.
Documentation seems to be outdated, wrong, or non-existant. I feel lost.

Is there *any* way to make this setup work again reliably or is there no
other solution than to switch to a less fragile software?!

Sincerely.

-- 
10 GB Mailbox, 100 FreeSMS/Monat http://www.gmx.net/de/go/topmail
+++ GMX - die erste Adresse für Mail, Message, More +++

Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html