On 9/22/2010 9:26 AM, Piotr Antczak wrote:
> Hi,
> I'm wondering whether it is possible to cluster dir and sd? Lets 
> assume that I have to nodes: node01 and node02. Each of this nodes can 
> access disks which are used as the file storage, can access a disk 
> devices on which catalog database will reside. Each of this nodes can 
> access SCSI tape library (both of them can access tape drives and 
> robotic arm). I think that I can do this this way (correct me if I'm 
> wrong) assuming active standby configuration:
> 1. Catalog database if configured as a cluster service with it's own 
> ip cluster address - storage for catalog database is accessible from 
> both nodes. Configuration files for catalog database are on nodes 
> local disk (and are identical) or the configuration files are placed 
> on common disk storage
> 2. director is configured as cluster service and it has it's own ip 
> cluster address. Configs can be placed on local or shared disks
> 3. storage daemon is configured as cluster service and it has it's own 
> ip cluster address. SD can access tape library and disk drives used as 
> File devices on each node (the mount points and scsi device names for 
> tape library are the same on each node).
>
> Is there a chance that it will be working?
>
The database and Dir work well. The SD is the tricky part. You must have 
a tape drive on a shared SCSI bus, such as with Trivoli Storage Manager, 
or some other means of sharing the tape library that can also provide 
for a SCSI bus reset on failover. I have SD running as a cluster service 
on a two node Linux corosync/pacemaker cluster using DBRD disk devices 
for shared File storage, which is an alternative that works in a 
disk-disk-tape setup, where the tape drive does not have to be clustered 
in a failover compatible way. In other words, the backup jobs write to 
shared disk with failover, then copy jobs copy the jobs to tape without 
failover, because the tape drive is only attached to one node. If the 
node the tape drive is attached to fails, then the tape drive can be 
manually moved to the other node and the copy jobs restarted. The copy 
to tape is not so critical, though, since the backup to disk is the 
"real" backup.

The only way to do tape failover is with a shared SCSI bus, AFAIK. In 
theory, it should work with a shared SCSI bus system.


------------------------------------------------------------------------------
Start uncovering the many advantages of virtual appliances
and start using them to simplify application deployment and
accelerate your shift to cloud computing.
http://p.sf.net/sfu/novell-sfdev2dev
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users