<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
<br>
<pre class="moz-signature" cols="72">Justin Bennett
Network Administrator
Dynabrade, Inc.
8989 Sheridan Dr.
Clarence, NY 14031
</pre>
<br>
<br>
Sam Stern wrote:
<blockquote cite="mid002301c664b8$cdf25a50$0bfe2c81@samstern.net"
type="cite">
<blockquote type="cite">
<pre wrap="">-----Original Message-----
From: Justin Bennett [<a class="moz-txt-link-freetext" href="mailto:Justin.Bennett@Dynabrade.com">mailto:Justin.Bennett@Dynabrade.com</a>]
Sent: Thursday, April 20, 2006 02:46 PM
To: <a class="moz-txt-link-abbreviated" href="mailto:samstern@samstern.net">samstern@samstern.net</a>; <a class="moz-txt-link-abbreviated" href="mailto:nflug@nflug.org">nflug@nflug.org</a>
Subject: Re: [nflug] NFS noac option
Thanks, I'll keep it in mind. We'll see if I run into any
issues with the hard mounts. I've set it with no options at
the moment we'll see how it goes. We also run maildir.
We've been running with the 'soft' option for years, It was
added because the machine wouldn't shut down during a power
outage before the UPS died, hanging on unmouting drives. I
think origionally my problem was with courier still having
file locks when the machine is going down. A while back
changed my shutdown script run by the UPS to shutdown
sendmail, courier imap, samba, autofs, then the nfs services,
</pre>
</blockquote>
<pre wrap=""><!---->
<SNIP>
Ahh. Okay I use Courier-MTA. The problem with Courier-IMAP and NFS (or iSCSI
for that matter!) is two fold:
1) Famd is used to maintain a list of read items. So make a point of
shutting that down as well -- perhaps having famd still running is causing
your problems?
2) Courier uses caching of keywords, acls and the like so these files can
become corrupted during an unclean shutdown
I've found that it's easiest to merge this into your recovery scripts before
courier starts to minimize problems:
Find $home/Maildir/ -type d -name courierimapkeywords/ -exec rm -dfr {} \;
Repeat for each of these files (putting these into a regexp made the email
look funny so I did not):
courierimapuiddb
courierimapuiddb.lock
sqwebmail-curcache.db1
sqwebmail-curcnt
This will force a re-read of the directories after an unclean shutdown by
the imap clients. The forced reread can seems to solve a problem with
Outlook seeming to get any new mail after an unclean shutdown.
Please note that should never delete courierimapsubscribed for obvious
reasons ;> This file can become corrupted in rare circumstances (like making
a new folder JUST when the server crashes). This has oddly happened to me
twice in the four years I've been using courier-mta. If the file becomes
corrupted due to a unclean shutdown, edit the file and remove any lines that
are corrupted. The file is not terminated by a new line; adding one has
caused me problems in the past. The last character in the file is always a
"." here is a sample for reference:
mail:/home/sam/Maildir # hexdump -Cv courierimapsubscribed
00000000 49 4e 42 4f 58 0a 49 6e 62 6f 78 2e 53 65 6e 74
|INBOX.Inbox.Sent|
00000010 0a 49 6e 62 6f 78 2e 44 72 61 66 74 73 0a |.Inbox.Drafts.|
0000001e
Further, take a look into this faq:
<a class="moz-txt-link-freetext" href="http://www.courier-mta.org/?imapd.html">http://www.courier-mta.org/?imapd.html</a>
Note the information on IMAP_USELOCKS
_______________________________________________
nflug mailing list
<a class="moz-txt-link-abbreviated" href="mailto:nflug@nflug.org">nflug@nflug.org</a>
<a class="moz-txt-link-freetext" href="http://www.nflug.org/mailman/listinfo/nflug">http://www.nflug.org/mailman/listinfo/nflug</a>
</pre>
</blockquote>
<br>
Thanks for the info. I;ll hold on to it incase I run into any
issues.I'll hold on it it incase I have a problem. We've been running
Courier maybe 3 years now with no problems. Maybe twice a year we have
a power issue which leads to a UPS triggered shutdown, but since I
changed it to shut all the relevant services down before NFS (using
soft mounts) I haven't had a problem. On recomendation from I've
changed it to run with the default options (with automounter) I'm
hoping that shutting all the services down, then NFS, then doing the
init 0 will be still go ok with the hard mounts, and not hang at
unmounting filesystems like I've had in the past (I was just doing an
init 0, not shutting anything down in a particular order, just whatever
the order was in /etc/rc.x). We'll find out next power outage.. :)<br>
<br>
<br>
Thanks<br>
Justin<br>
<br>
<br>
<br>
</body>
</html>