Thank you for attending Symantec webcast Take VMware vStorage APIs to the Limit with Unrivaled Data Protection. The recording of this webcast will be available soon. I am documenting the questions asked during the webcast and respective answers below. The names are withheld for privacy reasons.
Answers to questions asked during the webcast "Take VMware vStorage APIs to the Limit with Unrivaled Data Protection"
Created: 15 Nov 2012 • Updated: 19 Nov 2012 • 1 comment
1. We're seeing snapshots not being removed after the backup operation is finished. Issues?
NetBackup’s bpfis program is executing creation and deletion of hypervisor level snapshots using VADP (vStorage APIs for Data Protection) APIs. The snapshot deletion issues occur mainly because of hiccups within the ESXi stack. For example, the backup was too long (or there were too many snapshots at the same time) that resulted in near 100% space consumption. The best way to verify what had happened is to look at vSphere client for the events associated with the virtual machine(s) in question. The error related to snapshot deletion will be reported here.
NetBackup handles these kinds of unavoidable issues using Orphaned snapshot handling attribute in the policy. Although the problem cannot be prevented (because it is beyond the control of backup application) NetBackup does try to clean up those orphaned snapshots.
I would recommend working with Technical Support to look at the specific issues you have in your environment.
2. When setting preferred order of IO, ...ie, SAN, NBD, etc how can you see if SAN transport is indeed happening?
That is a great question! Currently, the only easy way to validate that a specific transport mode is functional is to deselect everything else. Unfortunately, the selected transport method is not logged in Activity Monitor. There is a way to identify this from VxMS logs on the backup host, but I would not recommend it as these types of logs are mainly for diagnostics. It can use up a lot of space if left unattended.
We have an enhancement request in place to bring the selected transport to Activity Monitor. Stay tuned.
3. Can you explain difference between SuSE linux and SuSE Linux for VMware?
Not much difference from functionality point of view. See this blog for an explanation in the context of NetBackup for VMware.
4. How can I re-play this webcast? Can someone send me the link?
Since you were registered, you will get an e-mail from our events team with a link to recording. It typically takes a few business days. I shall also update this blog when that recording is published.
5. Is there any reason we cannot get this level of performance if we are not using Cisco UCS?
The goal of this benchmark is to evaluate the performance characteristics of VADP. Obviously, we need to choose the infrastructure such that ESXi will not be the bottleneck while studying the behavior of NBD that uses VMkernel. These powerful Cisco UCS blades helped us to achieve this. You may be able to get similar level of performance with other servers as well. But please do note that your mileage may vary.
6. Can we use NetBackup NBD for backing up vBlock?
Absolutely! vBlock is very similar to FlexPod we used in this benchmark. The storage in vBlock is from EMC. The benchmark we did here is infrastructure-agnostic. So long as you are using vSphere 5.0 or higher, the best practices discussed here is applicable for any form of its implementation including vBlock.
7. What do you recommend for NFS datastores?
Preferred method for NFS based datastore is NBD transport from a physical backup host (for a ready-made solution, consider NetBackup 5220). Follow the best practices discussed in the webcast (limit 4 streams per ESXi, use VMware Intelligent Policy etc.). Note that NFS storage itself is served via VMkernel. It makes more sense to offload some of the workload from ESXi (e.g. deduplication) by letting the backup server do the heavy lifting.
8. Is there any limit recommended for san transport?
No theoretic limit for SAN transport. The bottle neck possible in the data path for SAN transport backups are a) The read performance of datastore b) bandwidth of FC or iSCSI links c) The write performance of backup storage. Identify the bottleneck and see if you can tune it further.
9. How do we get a copy of the slides?
You will get a e-mail with the links to recording and slides.