Veritas-bu

[Veritas-bu] Database Cluster Policy Strategy

2007-03-21 22:10:14
Subject: [Veritas-bu] Database Cluster Policy Strategy
From: mklein987 at gmail.com (Michael Klein)
Date: Wed, 21 Mar 2007 22:10:14 -0400
I'm curious how people manage database backups in a an active-active
or N+1 cluster in NetBackup.

I have four 2-node active-active SQL Server clusters (8 unique SQL
instances) that share the same backup parameters and same backup
window. The layout is as follows:

Cluster1
  Node1 hosts SQLVS1\Instance1
  Node2 hosts SQLVS2\Instance2

Cluster 2
  Node3 hosts SQLVS3\Instance3
  Node4 hosts SQLVS4\Instance4

?. you get the picture.

I'd like to configure a single policy to backup all 8 SQL instances. I
created a .bch file for each instance with a generic name
(SQL_Full.bch) and copied it to the dbext\mssql directory of the node
that is primary for each instance. My policy includes all 8 virtual
SQL server names and calls SQL_Full.bch. When all nodes of the cluster
host their primary instance (i.e., all nodes are online) everything
works perfectly. However, when one node of a cluster fails or is taken
down for maintenance, the other node hosts both SQL instances. In that
case, the primary instance is backed up twice and the instance that
failed over is not because the batch file on the single working node
(that now hosts both SQL virtual server names) only contains the
details about it's primary instance.

I've thought about putting the .bch file on each SQL instance's data
LUN and then referencing it in the policy but alas, there is no
standard drive letter among the 4 clusters (one uses X: & Y:, the
other V: & W:, the third T: & U:, etc.).

How do others manage database backup policies on multi-node clusters?


<Prev in Thread] Current Thread [Next in Thread>