[Mailman-Users] Internal Server Error

[Richard Barrett]

At 07:22 04/09/2002 -0400, Barry A. Warsaw wrote:

> >>>>> "JC" == Jon Carnes <jonc at nc.rr.com> writes:
>
>     JC> Then for a real shot in the dark, ask about the servers use of
>     JC> NFS and if any of its volumes are mounted remotely. There is
>     JC> nothing quite like trying to lock an application via an NFS
>     JC> mounted drive!
>
>Mailman's locking algorithm /ought/ to be NFS safe...
>
>-Barry

I concur that MM and NFS will mix OK.

I've been running our production MM (now at 2.0.13) with all MM related 
files, except those few installed in EXEC_PREFIX, in an NFS file system 
mounted from our Auspex file server. Why? Because our main file server has 
full disk mirroring etc and backup. It also lets me switch MM service 
provision to another backup machine in a few minutes with minimal manual 
intervention. Thus both Mailman and the Apache server on the MM server 
machine when handling MM data are doing so via NFS.

The only problem I've encountered with this use of NFS is a result of, I 
believe, deficiencies in the Linux NFS client on our MM server machine; I'm 
using Suse 7.3 on that machine but Redhat also exhibited the same problems. 
I found that NFS file accesses would hang on large data transfers to/from 
automounted NFS file systems; with Redhat 6.2 the system locked up 
completely requiring hard shutdown and reboot; Suse continued to run with 
the NFS transfer hung but the system could be rebooted normally. Explicit 
mount of the necessary file systems (from /etc/fstab at boot time) with 
read and write block size limited to 8096 seems to prevent the problem.