commvault.com commvault.com
Solved

VM index restore

  • 23 October 2023
  • 11 replies
  • 138 views
L
Rank
Badge +6

Hi, 

the customer has two locations. A and B. On both we have separate media agents with library attached. Media agents between locations don’t have any communication and we can’t change it.

He was doing VSA VM1 backup on location A. 

They have migrated VM1 to location B and started backup. Full backup is working fine but during incremental we have problem with restoring indexes. During incremental backup the job tries to restore index from location A to the index cache on media agent B. As there is no communication between them we’ve got an error and the job failed.

 

What can we do about it? Can we somehow recreate index on media agent B? 

Regards,

Lukas

 

icon

Best answer by Rajiv 25 October 2023, 12:17

View original

11 replies

J
Rank
Userlevel 1
Badge +7

Hi Lukasz,
did you try to define a network topology between the MediaAgent in A to the MediaAgents in B ?
they could use a proxied communication, where either the commserve or any other existing commvault gateway, capable to connect to both sides, is used.
or even a cascaded gateway is usably 
SiteA MA → GW1 → GW2 ← MA SiteB

rgds
Klaus

L
Rank
Badge +6

Hi,

the problem is that there is no communication between these two media agents and we can’t set it up. 

The customer resolved the issue. He has deleted all backups of VM1 which was left on site A.  

I am wondering why during incremental backup on site B it was trying to restore the index from site A even if we did full backup before. 

In this scenario, we could delete backups but what to do when we have to leave old backups on site A for a long time?

 

Regards,

Lukasz

 

J
Rank
Userlevel 1
Badge +7

Hi,
Commvault requires the client index to be present on a mediaagent associated with the storage policy used to backup the client.

An incremental backup allways requires an exsiting index to add to….

in case the mediaagent is changed due to load balance or storage policy change, the exsiting index is restored / transferred from the formerly used to the new media agent.

you can also observe this behaviour in case you have multiple mediaagents associated to the same storage policy (multiple datapaths).

nevertheless I assume, that at least you commserve is able to communicate to both sites.
which enables you to send data between sites using the commserve as proxy/gateway.

although this should not be used to send backup/restore data accross sites 

rgds
Klaus

L
Rank
Badge +6

Hi,

do you know if there is any way to re-create an index on a media agent without restoring it from backup? Just start a new index chain? 

Regards,

Lukasz

J
Rank
Userlevel 1
Badge +7

Hi,

to recreate the index, the mediaagent requires access to the storage policy where the former backups are stored.

ddi you try to run a fresh Full (not SynFull) backup on the affected VM only.
This might help (even though not sure).

A full backup starts a new cycle, which might bhe sufficient in your case.

nevertheless I’d try to solve the communication issues, since the initial scenario described can bring some instability to the backup solution

Restores of backup data stored to SiteA cannot be restored to a VM now running in SiteB. Hmm.

Are you sure this is a valid / wanted outcome ?

rgds
Klaus

L
Rank
Badge +6

Hi,

As I’ve mentioned before full backup was performed without any issues but following incremental was failed. In the logs, I saw that it was trying to restore indexes during incremental. Full backup doesn’t resolve the issue.

Regards,
Lukas

J
Rank
Userlevel 1
Badge +7

Hi, Lukasz,

can you verify the Indexserver chosen/defined in the backupset of the VMinstance ?
VM Client → Virtual Sever → VMInstance → defaultBackupSet

is this pointing to a mediaagent on the desired location ?

L
Rank
Badge +6

Hi,

yes, it is pointing to the media agent/access node on site B. 

Regards,

Lukasz

Rajiv
Rank
Userlevel 4
Badge +11

Hello @Lukasz A the index checkpoints might still be present on the old Index MA and hence when an incremental job is run, its trying to fetch it from there and it does not exist and it fails. 

Have you deleted all the backups of the VM1 and the data pruning has run against that Storage Policy?

If there are no old jobs against VM1 left in job history and pruning has also been completed and still you are getting the incremental backup issue, please raise a case with us since we might need to delete the reference check points from the DB.

Best,

Rajiv Singal

L
Rank
Badge +6

Hi @Rajiv,

as I’ve mentioned above when the customer deleted backups from site A incremental backup started working on site B.

The question is what if we can’t delete old backup because of for example internal regulations? 

Regards,

Lukasz

Rajiv
Rank
Userlevel 4
Badge +11

@Lukasz A In that case you can either establish the communication between both the sites so that Indexes can be restored or you raise a case with us to further validate if the issue is regarding Index checkpoints and we can remove those check point entries pointing to old MA.

Best,

Rajiv Singal

Reply


Terms & ConditionsCookie settings