Veritas-bu

[Veritas-bu] Question in regards to high availability of the catalog.

2006-03-17 17:30:10
Subject: [Veritas-bu] Question in regards to high availability of the catalog.
From: peter_marelas AT symantec DOT com (Peter Marelas)
Date: Sat, 18 Mar 2006 09:30:10 +1100
Justin,

I have setup cross-site NetBackup Master failover using the VERITAS SFHA
stack (VxVM/VxFS/VVR/VCS/GCO).

Sites are ~40 km apart.

VxVM is used to virtualise the storage and provide online storage
management and snapshot.

VxFS is used to provide file system journaling.

VVR is used to replicate the NBU binaries and catalogues asynchronously
over IP.

VCS provides local failover and integration with GCO.

GCO provides wide area failover and interfaces with DNS.

In the event of a site failure GCO triggers a service group failover
which updates the DNS, promotes the VVR secondary to the primary
replication role, fsck's and mounts the disks and starts NBU services.

On failback, we perform a manual sync (depending on why the production
site failed in the first place) and then trigger a GCO site failover.
The same process applies.

It works.
It is automated.
It fails over in under 5 minutes.

Environment is NBU 5.1.

Notes

- For VVR to replicate, storage must be block based. In your case Netapp
FC or Netapp iSCSI.
- You need to roll your own NBU start/stop/monitor/clean scripts for VCS
(not hard).
- Your DNS needs to support dynamic updates. GCO can update multiple DNS
servers.
- Your DNS service needs to be highly available and DR capable.
- All your clients and media servers must reference the NetBackup Master
using DNS.
- The DNS TTL for the Master record must be low (we used 60 seconds).
- Keeping the NBU binaries on replicated storage reduces operational
burden.

Regards
Peter Marelas


-----Original Message-----
From: veritas-bu-admin AT mailman.eng.auburn DOT edu
[mailto:veritas-bu-admin AT mailman.eng.auburn DOT edu] On Behalf Of Justin
Piszcz
Sent: Saturday, 18 March 2006 2:14 AM
To: Paul Keating
Cc: NBUList (E-mail)
Subject: Re: [Veritas-bu] Question in regards to high availability of
the catalog.

I guess the question is can we have each mirror of the catalog running
"hot" between both sites?  So masterA dies and then masterB is booted
and takes over masterA's DNS entries and then continues to use the
catalog which is a live mirror.

Secondly, is there anything in NetBackup 6.0 in regards to high
availability over the 3/4/5 versions?

The big question is what type of fail-over systems exist in NetBackup?

Is there anything that comes close to an Oracle cluster fail-over?

Justin.

On 3/17/06, Paul Keating <pkeating AT bank-banque-canada DOT ca> wrote:
> Shouldn't make a difference.
>
> From what I understand, you want to run your catalog backup with the 
> destination set to a disk path that is an NFS mount on your Netapp, 
> then have the NetApp replicate to your second site?
>
> No reason not to try it.....Netbackup shouldn't care, as it believes 
> it's writing a file to disk....the replication to a remote DataCenter 
> is irrelevant to Netbackup, as it's only "writing" to the local
filer...
>
> If it were me, I would set up the regular catalog backup so that one 
> destination was a tape, and the second destination was the filer
mount.
>
> I've been thinking of doing something like this myself....I have 
> various means of replication available....
>
> Currently using VVR, to replicate /opt/openv to the standby NBU server

> at the other location.
>
> Things I've been considering (instead of VVR) are:
> Configure catalog backup to point to a tape in the remote site's 
> library (linked via FC anyway, and used for offsite tape copy) 
> Configure one of the catalog backups to point to an NFS share mounted 
> from standby server at remote site.
> Configure one of the catalog backups to point to an NFS share on local

> NetApp filer which gets replicated to remote filer.
> One or more of the above + put catalog on SAN (not NAS), which is 
> synchronous replicated to remote site.
>
> Paul
>
>
>
>
>
> > -----Original Message-----
> > From: veritas-bu-admin AT mailman.eng.auburn DOT edu
> > [mailto:veritas-bu-admin AT mailman.eng.auburn DOT edu] On Behalf Of Justin

> > Piszcz
> > Sent: March 17, 2006 6:49 AM
> > To: Mahesh Ravindra
> > Cc: NBUList (E-mail)
> > Subject: Re: [Veritas-bu] Question in regards to high availability 
> > of the catalog.
> >
> >
> > Mahesh,
> >
> > Understood, however my plan is to perform an incremental catalog 
> > backup say once a minute with syncs occuring once a minute also so 
> > the catalog is "live" in two separate datacenters.  The idea behind 
> > it is if datacenter1 is lost for any reason, then we can install 
> > netbackup master software on the fail-over server, run bprecover 
> > with the path to the catalog  on the netapp and then be good to go.
> >
> > Have you seen or do you have any information or documentation 
> > regarding running catalog backups and mirroring them across multiple

> > sites?  While I have not read the HA document yet, from what I 
> > gather, they assume both the storage and the remote server are 
> > ``local'' to the installation and not across multiple sites.
> >
> > Thanks,
> >
> > Justin.
> >
> > On 3/17/06, Mahesh Ravindra <mravindra AT apara DOT com> wrote:
> > > Hi,
> > >
> > > netapp does snapmirror based replication of catalog u can
> > schedule the
> > > same in such a way that it gets replicated once in 24Hrs.
> > >
> > > and it works on snapshot based technology,we have been
> > doing the same for
> > > customers using netapp .
> > >
> > > Thanks
> > > Mahesh
> > >
> > >
> > > > Has anyone performed a high availability catalog across 
> > > > different datacenters?
> > > >
> > > > For instance:
> > > >
> > > > Master [Datacenter1]  --------------> Netapp [Datacenter 1]
> > > >                                                   ^^^
> > > >                                                   |||||
> > > >                                                   vvv Master 
> > > > [Datacenter2 (standby)] -> Netapp [Datacenter 2]
> > > >
> > > > Where the two netapps are synced via their own protocol
> > for replication?
> > > > Has anyone done this or similar methods of catalog
> > replication across
> > > > multiple sites?
> > > >
> > > > Thanks,
> > > >
> > > > Justin.
> > > >
> > > > _______________________________________________
> > > > Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu 
> > > > http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
> > > >
> > >
> > >
> >
> > _______________________________________________
> > Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu 
> > http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
> >
>
> _______________________________________________
> Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu 
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
>

_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu