Restore slowness

[smoon74]

I have a question regarding the slowness of Veeam. I'm attempting to use Veeam BR to restore a vmdk file, but it's extremely slow to copy 50GB ~ 4-5 hours. The Veeam vbk file is located on my laptop, and I'm attempting to restore to a datastore located on an ESXi host on the same network. But, my vCenter server is located at our main site, which is across a T1. So, I suspect the slowness is due to the fact that when I run Veeam BR on my laptop Veeam communicates across the T1 to vCenter.

Does anyone have any suggestions on the best method to restore a vmdk and associated files. Is it best to use FastFCP?

[Gostev]

FastSCP and Veeam B&R is the same thing (FastSCP has reduced/free functionality). You may also want to try adding your ESXi as standalone host (for example, by IP address), to remove the need of communication with vCenter.

Generally speaking though, file copies to ESXi can be quite slow indeed. You can try uploading a file via vSphere Client Datastore Browser, and see for yourself. Most typically, the speed reported is about 5-10MB/s (18-36GB/hour). I believe vSphere 4.1 improved this performance somewhat, but not 100% sure on this.

[ThomasMc]

The best way I've found to do a "Full Restore" is to publish the VM via Instant Recovery but don't power on(unless your lic allows you to Storage vMotion) after its been published right click the VM and choose Migrate -> Change datastore -> Choose a LUN(not the Veeam one of cource) -> Choose your Disk Format -> Next and then Finish

You may get a couple of errors like

VM already registed(just remove from the inventory)
and
Files already there(choose a diffrent lun or delete the VM from disk that your trying to fully restore)


My test was;

30GB VM via Veeam "Full Restore" option was 4hrs to complete and the above had the VM up and running within 8 mins(Storage vMotion would have reduced that to seconds)

[Gostev]

Thanks Thomas - yes, that might be a good workaround indeed, what helps here is the fact that data is copied differently - using host's internal, highly efficient data mover. On top of that, I can add that we also have something in the works that should hopefully make it to the next release and allow for faster restores w/out Instant VM Recovery workaround. Thanks.

[ThomasMc]

That would be awesome, its the only thing that I can find wrong with Veeam B&R(not your fault I must add)

[tedsteenvoorden]

[merged into existing discussion]

We are currently implementation Veeam Backup & Replication 5.0.2 and are testing the functionality of the product. We currently backup our VM's via the direct SAN method. Our VMware environment is based on ESXi 4.1 update 1. Testing the restore functionality, we noticed the restore speed of a full VM is very slow. I found out that restore via direct SAN is not supported and instead the network is being used. Our ESXi 4.1 update 1 hosts are connected through a 1 Gbit network but the restore rate is around 10 MB/s.

Reading trough this forum i am getting the idea this is an ESXi issue but i am not sure. Can somebody report if this is expected behaviour?

[Vitaliy S.]

Yes, you're getting it right, the same situation you would get if you try to upload any file to ESXi host via Datastore Browser.

[tedsteenvoorden]

Will their be a different in speed using ESX 4.X with a Veeam agent instead of the agentless restore with ESXi 4.X?

[Gostev]

Yes, the speed will be much better. Thanks.

[SarahCurtis]

[merged into existing discussion]

Hi

We are still running ESX and are seeing very fast restore speeds. When testing with ESXi, we are seeing very slow restores due to the lack of a Veeam agent in the service console. Given that VMware are forcing us down the ESXi route, what plans do Veeam have to get around this slow ESXi restore issue? And, if plans exist, when can we expect a solution to be made available?

Thanks in advance.

[ThomasMc]

This was my workaround

The best way I've found to do a "Full Restore" is to publish the VM via Instant Recovery but don't power on(unless your lic allows you to Storage vMotion) after its been published right click the VM and choose Migrate -> Change datastore -> Choose a LUN(not the Veeam one of cource) -> Choose your Disk Format -> Next and then Finish

You may get a couple of errors like

VM already registed(just remove from the inventory)
and
Files already there(choose a diffrent lun or delete the VM from disk that your trying to fully restore)


My test was;

30GB VM via Veeam "Full Restore" option was 4hrs to complete and the above had the VM up and running within 8 mins(Storage vMotion would have reduced that to seconds)

But Gostev said there was already plans to implement a fix in newer versions

[SarahCurtis]

Thats good news, just wondered what the ETA was on this, given that we have ESXi offsite and are seeing very slow DR testing as a result, do we swap all this out for ESX to meet our SLA's, or do we wait given that a 'fix' from Veeam will be available in x days/months...

[Beevoir]

Hi Sarah,

Are you backing up over the WAN or Replicating over the WAN? This can make a difference to the restore time and also the transfer times depending on which side of the WAN the Veeam server is on.

Also, for DR testing, have you seen the Surebackup functionality within the Veeam product? It does all of your backup DR testing for you!

Mike

[ThomasMc]

Wouldn't the Instant Recovery meet your needs in the interim? I can have a backup image loaded and live in about 30 seconds

[SarahCurtis]

I'm aware of the instant restore workaround. My issue is really to do with planning - we backup locally and can do SureBackup restore testing locally, but then replicate offsite. Once backups are offsite a 3rd party has the responsibility of doing biannual dr testing, and their estate is ESXi. We need to know if its worth asking them to install some ESX just for us, which they understandably don't want to do given the VMware roadmap, or whether its worth waiting as Veeam will be providing a fix in x weeks/months. Just after some sort of estimate as to when this won't be an issue any more...I know this is a VMware problem, but Veeam have hinted a solution is in the pipeline...

[ThomasMc]

Maybe I'm being a bit dim here but I don't see the point in that, I would expect your DR site to have Full Enterprise License to enable them to Storage vMotion so their recovery should be something along the lines of;

1. Publish the VM via Instant Recovery and power on
2. Storage vMotion to a live LUN
3. go make a coffee

[Added]
Oh I think I get you now, is it getting the VM from your DR site to your Production site thats in question?

[Vitaliy S.]

...whether its worth waiting as Veeam will be providing a fix in x weeks/months.

The functionality you're referring to will be shipped with our new major version, no "fixes" will be available for the current release.

[stormlight]

Am I correct to say that this work around assume that you have Vspehere virtual center installed on the DR site. It will not work if you just have a single ESX host at DR. You dont get a migrate data store option without vsphere vc

The best way I've found to do a "Full Restore" is to publish the VM via Instant Recovery but don't power on(unless your lic allows you to Storage vMotion) after its been published right click the VM and choose Migrate -> Change datastore -> Choose a LUN(not the Veeam one of cource) -> Choose your Disk Format -> Next and then Finish

You may get a couple of errors like

VM already registed(just remove from the inventory)
and
Files already there(choose a diffrent lun or delete the VM from disk that your trying to fully restore)


My test was;

30GB VM via Veeam "Full Restore" option was 4hrs to complete and the above had the VM up and running within 8 mins(Storage vMotion would have reduced that to seconds)

[ThomasMc]

Am I correct to say that this work around assume that you have Vspehere virtual center installed on the DR site. It will not work if you just have a single ESX host at DR. You dont get a migrate data store option without vsphere vc

That would be correct but I would assume that the VC would be over in the DR site with the rest of your VMs and be one of the first VM to be Instant recovered(but manually moved once the rest of the site is up). If there is no VC on the main site or the DR site then I don't see it working, replication to the single ESXi host would probably more suit but that doesn't sound like what Sarah is doing

[Beevoir]

Hi Sarah,

The official answer for the release date of Veeam BaR is during Q4 of this year.

Do you replicate the backups offsite or are you using the replication engine to take them off? The reason that I ask is that you are able to do one of two things with this model.

Because of the way that we license, aside from the resources for an additional VM on the remote site you are perfectly entitled to have a secondary Veeam server. If you are using storage mirroring/LUN transfer for the backups themselves, you can script the importing of a backup file into the Veeam server on the secondary site and run a surebackup job there.

Or, if you are using the Veeam replication engine, you can script the startup of each VM you replicate (because of the way that we use reverse incremental to take the replica) and test the machines that way.

I hope that gives you a couple more ideas.

Thanks

Mike