Cyrus, clusters, GFS - HA yet again

[Dave McMurtrie]

Janne Peltonen wrote:

>>If you decide not to pursue a cluster solution, Perdition would probably 
>>help you with this part.
>>    
>>
>
>Ok. Is there any gain in using Perdition instead of Murder? Is it more
>stable? Less complicated? More widely used? Better suited to the system
>of our size (why)?
>
>  
>
I wouldn't be able to compare these for you. I only recommended 
Perdition because I know that it exists, but I have never used it.

>So I'm confused. I /am/ using a clustering FS with correct locking and
>memory mapping, but I'm still getting the (apparent) DB errors Chris and
>others said there would be. That's why wanted to ask those that have
>succeeded in creating a clustered Cyrus - be it on Veritas or the Tru64
>cluster system - how did they do it.
>
>  
>
I can only speak from experience.  When I worked at University of 
Pittsburgh, we set up Cyrus on Veritas Cluster and it just worked.  The 
only changes I recall that we had to make were:

1. We added a "nodename" config option for the sole purpose of making 
sure the lmtp temporary delivery files would be unique across the cluster.
2. We had some performance problems caused by the way that Veritas 
Cluster Filesystem dealt with mmap()d files.  We ended up modifying the 
Cyrus code in map_refresh() such that it would MAP_PRIVATE instead of 
MAP_SHARED to deal with this.  Veritas has since fixed this problem, so 
this change would no longer be necessary.

Other than that, we didn't have to go through any heroics.  We set up 
Veritas Cluster on 4 Sun V440s running Solaris 8 connected to an EMC 
Clariion SAN, then installed Cyrus imap.  There was nothing else to it.  
It was very simple and it performed very well.  I no longer work there, 
but to my knowledge it's never been down.

Thanks,

Dave