Re: [Linux-cluster] where should i look for an answer (postfix on gfs)

[Greg Swift <gsml netops gvtc com>]

Hirantha Wijayawardena wrote:
> Hi
>
> I have 2-node postfix mail cluster and I didn't mount /var/spool/postfix on
> the storage though /var/spool/mail. If fail-over has occurred, manually cp
> queues to the other node and $ postsuper -r ALL - Of course I may write a
> script to cp and do the postsuper.  
>
> Hope this is help and keep me posted if you found a solution.
that is helpful, always good to look at it from different angles. I'll 
keep in mind the queue option (its kinda what i was aiming for, but with 
the data on a shared storage, so if the box died copying would be 
easier) but I dont have much in the local /var/spool/mail since this is 
strictly an outbound relay (the reason why the queues are rather 
important). Another reason I was trying for the gfs share was that in 
theory the fibre attached storage has a better capacity for speed than scsi.

After some troubleshooting on the postfix list this was the latest:

> > Nov 28 21:34:15 rushou logger: write(5, "\260P\0\0\1\0\0\0\0\0\0\0", 12) 
> > = 12
> > Nov 28 21:34:15 rushou logger: read(6, 0xbff36a98, 1024) = -1 EINVAL 
> > (Invalid argument)
>   

This O/S (or FIFOs specifically) is broken. From Linux manpages:

   EINVAL fd is attached to an object which is unsuitable for reading.

Postfix needs working FIFOs.

Which when i suggested gfs2 might not be working the way i need it on my setup I received a suggestion that I might have the drive mounted as 'nodev', so i'm bring the system back up to check ryt now. And it does not appear to be.

thanks!

-greg

--
http://www.gvtc.com
--
“While it is possible to change without improving, it is impossible to improve without changing.” -anonymous

“only he who attempts the absurd can achieve the impossible.” -anonymous