commvault.com commvault.com
Solved

What are the Pros and Cons Clustered Media Agent or non cluster Media Agent

  • 24 February 2022
  • 7 replies
  • 645 views
K
Rank
Userlevel 2
Badge +4

We are planning to deploy a Clustered Media agent so need more details on advantages upon configuring a non-clustered Media agent.

icon

Best answer by jgeorges 15 March 2022, 09:33

View original
If you have a question or comment, please create a topic

7 replies

Onno van den Berg
Rank
Userlevel 7
Badge +19

@Kavya what type of backend storage are you planning to use? clustering the MediaAgent can be done, but comes with disadvantages like for example added complexity and the fact that it runs active/passive. only reason I can think of to cluster your MA is when you are limited to block storage alone, but even than I would still consider not using clustering and instead using two standalone MAs. 

K
Rank
Userlevel 2
Badge +4

We are using physical media agent as the primary storage hence we are planning to a clustered MA.                                    

Onno van den Berg
Rank
Userlevel 7
Badge +19

@Kavya are you planning to use storage that reside in the server that acts as the MediaAgent or are you attaching them to external storage using FC or network?

K
Rank
Userlevel 2
Badge +4

we are using its internal storage of the physical server with high Storage availability

J
Rank
Userlevel 5
Badge +9

Hi @Kavya 

If you’re using a Physical MA with internal disk, if the server goes offline, than the storage goes offline with it. I’m not sure how you maintain HA with internal disk storage?

If you’re using Deduplication, than thats catastrophic to your failover plan, as backups will be unavailable while the DDBs offline, and reconstruction will fail as the storage is unavailable for reading.
If you’re not using deduplication, than backups can continue to run to the ServerB Storage, however restores of the previous backups will be unavailable.

 

If the storage is accessible from both ServerA and ServerB, than this will be okay, with limited CONs outside of complexity and management (You’ll need to run updates on 2 servers at all maintenance intervals).
The DDB folder structure will need to be recreated on ServerB and the reconstruction job run, however this will run as the storage is accessible for reads, and once completed backups will continue without any loss to continuity.


Generally, i don’t think the media agent is to hard to ‘replace’. In the event of a failure, its more important to have good DDB Backups on accessible storage. If you do, than the time it takes to stand up a new MA is roughly the same as the time it takes to re-install MA binaries and ‘refresh’ the existing MA. 

If you did have 2 Physical Servers with internal disk, i’d say a good option would be to dash copy data between the two.
In the event of the ServerA failing, than you have recoverability from ServerB, as well, you can ‘promote’ ServerB to become the primary and continue running backups to it.

 

Cheers,
Jase

K
Rank
Userlevel 2
Badge +4

I understand, but due to budget constraints we are not opting for 3rd party storage.

So, my question is, does having this Clustered VM’s as the Media Agent solves the availability and redundancy in case of any failures that might affect the data.

J
Rank
Userlevel 5
Badge +9

Hi @Kavya 

You’ll need to give us some more detail around the design of the environment and how / what you’re using to implement this.

You’ve already said the storage is internal disk, how are you replicating the storage between the two physical nodes? 

If you’re using Microsoft Server Clusters and implementing Storage Replication, than presumably this would mean that the Physical Server at Site A is identical in every way, besides the hostname, to the physical server in Site B.
NB: i have very little understanding on how Microsoft Clustering and Storage Replication works, so someone else or perhaps Microsoft might be able to provide better feedback here

Being a Windows Cluster, the Media Agent is configured to leverage a psuedo name, which points to to the active node (managed by Microsoft clustering). Seeing as the storage and the servers are replicas,  during a failover, Commvault won’t know any better that the server has changed homes and continue to run as normal. 

There would likely be some downtime during a failover if the DDB was active, but this would be resolved with a ddb reconstruction. Indexes would be replicated and if required can again be reconstructed also.
 

Cheers,
Jase

Terms & ConditionsCookie settings