> Hi.
> 
> I have a cluster with shared storage which is mounted only on one node
> at a time. Other node's don't have that partition mounted. I wonder,
> what is a way to backup a shared partition?
> 
> If I only include partition in filesets of all cluster nodes, that
when
> storage is mounted on another node, incrementall will become full
> because in backup of that node, there aren't any files from those
> partition, because that partition was not mounted before... And I want
> to avoid that and possibly have only one backup job/location for those
> files.
> 
> One idea is to have a separate bacula-fd running as a cluster service,
> which will be started within the resource that mounts a partition.

That's how I'd do it, assuming you had a cluster ip address to that the
director could contact it on. That's how we got clustered Microsoft
Exchange backup working too - we made a 'cluster' bacula instance and
migrated it with the Exchange resource.

> Any
> more ideas, where bacula-fd does not need to be included as cluster
> service, but as a node service?
> 

Well you could just add it to the filespec on each node. If it's not
mounted then the directory is empty and you back nothing up. As you have
pointed out, this falls to pieces completely if you are using accurate
incremental/differential backups though as Bacula will deduce that all
the files previously in the directory prior to the resource migration
must have been deleted.

If it's a clustered resource then a clustered fd is going to be the best
way to handle it imho.

James


------------------------------------------------------------------------------
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users