02-20-2017 08:33 AM
Hi everybody!
We want to implement the SQL intelligent policy on our SQL cluster (VCS). But faced the trouble.
We have two different datacenters. SQL cluster deployed on 4 physical servers (two in each DC) and backup traffic between DCs is undesirable.
How can we set up the policy(-ies) to use storage unit from DC where SQL cluster instance is active at the certain moment during backup?
02-20-2017 09:08 AM
Create a Policy at each data center for local backup. Activate the Policy at the live site and Deactivate the Policy at the idle site. Don't know of any automated way to handle that.
02-21-2017 04:04 AM
Hello,
refer to "virtual storage units" topic in High Availability Guide (pages 22-27).
No matter if they are used with SQL intelligent policy or any other type of policies.
Regards
Michal
02-21-2017 04:23 AM
If memory serves me right, we have solved this issue like this (using traditional script-base policies):
Create 2 (or more, depending on number of nodes and sites) Application Backup schedules in the policy (e.g. siteA and siteB). Override the STU in each of the schedules to select either siteA or siteB Storage Unit.
Copy the backup script to local storage on each of the nodes.
Add NBSCHED variable to each script that specifies the Application Backup schedule with local STU.
Hope this makes sense!
02-22-2017 01:44 AM - edited 02-22-2017 01:44 AM
Many thanks to all for answers. I contacted to veritas support yesterday. They confirmed there no features at the moment for resolving this issue. And suggested to use scripts too.
02-22-2017 04:44 AM
As Michal suggested you can use the app_cluster feature in netbackup. Its usually used within a site but it would probably work across sites too. Failing that, you can put a preonline trigger into vcs to execute a policy storage location update prior to onlining the service group in each cluster.
02-22-2017 10:00 AM
02-23-2017 03:10 AM
Yes, agreed, SQL nodes must be configured as Media Servers for app_cluster functionality
Michal
02-23-2017 07:50 AM
Very nice idea. I did so over 10 years ago with a Fileservercluster (VCS) on solaris. The postonline trigger changed changed the storageuntigroups and kicked off the bpsched to be aware of these changes.
You should do something sililar with STU groups or SLPs. (bpsched is history)
good luck
tunix2k
02-23-2017 08:14 AM
As Marianne said, it should be a Enterprise Client/SAN Media Server for the app_cluster to work.
02-23-2017 10:29 AM