Question

Airgap: parallel job transfer from the secondary copy

  • 27 September 2023
  • 4 replies
  • 69 views

Badge +1

Hi Everyone,

Currently we are making traditional backups from HS to HS (Primary - Secondary Copy).
We added an extra Cloud Library as AGC location as mentioned in the picture below.

Goal
The AGC NEEDS to finish within 24 hours contains File System + SQL Agents that were backed up from the last 24 hours on the Disk Libraries.

Observation
From the {Secondary to AGC} it seems that new bckjobs are not picked up after the aux copy started.
Example: If bckjob 1-4 contains 10TB of data, commvault waits until these ones are finished and then will pick up the next bckjobs while some Media Agents in the Cloud Library are doing less work.

Conclusion
The {Secondary to AGC} takes to much time (+/- 48 hours), after the Full Backups are finished in the weekends on the Disk Libraries. It does NOT meet the requirements.

Questions
1. Are my observations correct if yes, how to tune it?
2. Any other options to meet the requirements?

 


4 replies

Userlevel 6
Badge +18

What speeds are you getting on both Aux Copy jobs?
How many streams are you getting on both the jobs?

Where I’m going with this … if the 10TB includes big Synth Full jobs, I’m curious if you’re using multi stream Synth Full jobs (requires Indexing v2) in order to increase the # of Aux Copy streams.  And, in general, more streams typically means more performance (within reason).

Note:  I’m assuming both of these are deduplicated copies, so make sure the DDBs are performing well and are not giving you a bottleneck.

Thanks,
Scott
 

Badge +1

From the example above, the 10TB of ‘backupdata to AirGap’ contains only SQL data.
Yes, those are deduplicated copies from the Secondary Copy.
The SQL-storagepolicy can take 25 streams but is never using 25 streams but only 10.

The current average speed for AuxCopies on the Disk Libaries +/- 4000GB/hour which is good.
The current average speed for AuxCopies to AGC is +/- 700 GB/hour.
But the thing here is, it starts for the first couple of hours with 1500 GB/hour and it slows down to 500 GB/hour. 
 

Userlevel 6
Badge +18

You might need a ticket with support to identify the bottle neck in the AGC aux copy jobs.

How are the resources (CPU, etc.) on the Cloud MA?  Are the DDB Q&I times good?  Are there any network bottlenecks?

Thanks,
Scott
 

Badge +1

The DDB Q&I is below “500” and we use “extra-large” MA’s from the Commvault Best-Practices sheet. Support was involved while setting up this configuration.

I was wondering if there are any members in the community with this kind of experience.

Reply