We're running qmail-ldap-1.03-20031001 on 4x Solaris boxes behind a
load balancer; they all have local LDAP replicas and deliver to a
shared NetApp NFS server. When we do a small stress test -- quickly
sending a few thousand messages to a local account via SMTP -- we see
hung qmail-ldap processes and stuck messages. It's been like this for
about three hours now:
qldap 17797 1723 10 07:53:21 ? 16:10 bin/qmail-local -- rdunbar
/nasahq/data/maildir2/rdunbar rdunbar newhorse.hq.
[EMAIL PROTECTED]> sudo qmail-qread
12 Dec 2003 12:53:20 GMT #9946 503 <[EMAIL PROTECTED]>
local [EMAIL PROTECTED]
[EMAIL PROTECTED]> sudo qmail-qstat
messages in queue: 1
messages in queue but not yet preprocessed: 0
Unfortunately, 9946 is used over and over again as it's the queue file's
inode (man qmail-log). So how do I correlate this to events in the
logs? There are 10 messages in my logs for a "9946"? This one has the
same timestamp:
2003-12-12 07:53:20.680988500 new msg 9946
2003-12-12 07:53:20.680994500 info msg 9946: bytes 503 from <[EMAIL PROTECTED]> qp
17795 uid 65026
2003-12-12 07:53:20.680998500 starting delivery 11199: msg 9946 to local [EMAIL
PROTECTED]
There is no other "11199" delivery message in the logs.
I also see this error, which is generated from maildir++; I don't
believe it's related to the stuck message above but occurs right after
the first logs for it, so I thought it might be related
2003-12-12 07:53:20.855999500 delivery 11197: deferral:
Warning:_undefined_mail_delivery_mode:_normal_(ignored)./Problems_while_trying_to_get_maildirsize:_file_already_exists._(QUOTA_#1.1.1)/
Have other folks using NFS for maildirs noticed issues? Any hints on
tuning?
Thanks.
