Number of VM's per job

[mcrdr85]

and the job from last night is still running... its been 14 hours and its only 81% complete!

[yizhar]

Hi.

I'm against huge jobs with many VMs, and this is regardless to the performance issue.
The main problems with large jobs are related to management and troubleshooting.
IMHO splitting to several jobs with 10-20 VMs each will give you better control.

Yizhar

[yizhar]

Hi.

Regarding performance - splitting the jobs and running 2 jobs in parallel (or using parallel processing) might help,
so you should try it.

Also try the following:
Install a physical server - for the initial test you can use a PC with 4 cores (for example i7 CPU), and 2 NICs.
Use it as proxy with 1 nic connected to LEFTHAND and configured for SAN mode backups.
If initial testing yields better performance, you can consider improving hardware further.

Yizhar

[foggy]

Also logs review could help in determining what operations take longer than needed, so I recommend opening a case with technical support.

[luckyinfil]

It should be pretty obvious that hotadd isn't the approach you want. Each backup requires approximately 5 minutes to hotadd the VMDKs and to remove the VMDKs afterwards, so when you scale that for hundreds of VMs, the majority of the backup window is spent mounting and unmounting VMDKs versus actual back up. If you need absolute speed, you will need a physical server and do SAN mode which bypasses the large amount of time used mounting and unmounting when using hotadd/virtual appliance.

[mcrdr85]

well i took half the VMs and put them into another job. It only took 2 hours!!! so i am going to just build a new job and put the other half of the VMs in it. Maybe something with the job is screwed up!

[Vitaliy S.]

To make sure it doesn't happen again, I would suggest reviewing the job log that took too long to process your VMs.