We are still struggeling with Agent for VMware to get reliable backups.
V-79-57344-38721 is still present (regression bug introduced with hotfix 164220) and unfixed. Symantec support suggested the workaround to backup VMs directly via virtual hosts and not via VCenter. This works however is really ugly if you have to handle a lot of VMs spread over multiple virtual hosts which are VMotioned from time to time. Symantec does not seem to put many effforts in fixing the issue.
Second problem we have:
When running multipe VMware agent backup jobs in parallel the backups very often hang infinitely. When trying to abort the jobs, abortion also hangs infinitely. You have to restart BE services to get things sorted.
Anyone elses having this problem?
For workaround we have scheduled the VMware jobs so they do not overlap, however this increases our backup windows noticably.
What's interresting is that if this occurs one of the VMware jobs is allways in the update cataloge state.
To me it seems like a multi threading deadlock problem or similar.
harsi: Have you checked the Known Issues section to see if this issue still exists? Have you logged a call with Symantec to see if they can assist you properly to fix the problem (not a hatchet job where your jobs are guaranteed to fail once a VM vmotions?
See this document for an explanation of what is done during "updating catalog".
It is more likely that excessive I/O's to the B2D folder cause a protocol mis-step.
Possible but unlikely in our case.
Why it only happens on multiple VMware agent jobs running simultaneously?
Doing the same for other jobs (MSSQL / Exchange / AD / AOFO) in parallel doesn't cause any issues.
Also the update catalog phase doesn't take very long and backup speed is quite acceptable.
When cumulating all backup jobs running in parallel we have a throughput of about 13.000 MB/s
There are a lot more things to ennumerate when you deal with a VM. Every file, every mailbox, database, AD object would have to be ennumerated. If you are having a big VM, then these objects can run into the thousands.
Yes but this doesn't match in our case.
We don't do application level GRT for VMs only file level GRT.
All BE supported apps are backed up the traditional way via remote agent even if they are running on a VM.