Restoring VMs with vStorage API

[clocatel]

Hello
I didn't find in guides if VEEAM backup could restore VMs on FC. Neither in the FAQ.
Could someone tell me the answer ?

Cedric.

[Gostev]

Hello Cedric, in the current version full image restores are not direct to SAN, they are performed through connection to the ESX host. Thank you!

[fgw]

anton, i already asked this in another thread, but as it perfectly fits here:

is veeam working on a solution to restore directly to san without going through an esx server?

dont know if this is even possible by the storage api?

would this be possible but it is not yet implemented, or would this not be possible at all?

[maxbbs]

Hi

I am trying to achieve a solution that can leverage LAN-free restore.

I am wondering if the following example configuration would work:

I have a 5 node ESX cluster with one shared datastore - Datastore1
I add another (backup) node to the cluster and give it access to Datastore1 via LUN masking etc.
I add a chunk of 'private' direct attached storage to this backup node. I create a datastore called 'backups' on this private storage.

So my backup node will be a part of the cluster the same as all the other hosts, but it will also have a large vmfs 'private' backups area.

If I then perform backups using the SAN option (with or without vStorageAPI), using the 'backups' volume as a target. This of course will be LAN free.

If I restore a VM from the 'backups' vmfs to the 'Datastore1' vmfs (is this possible on the same host?) - This will effectively be LAN free? Would all data be pushed through the (local) SC connection (still technically LAN free but possible big performance hit)?

If I restored a VM from the 'backups' vmfs to the 'Datastore1' vmfs but specify a different host in the same cluster during this operation, will this use the network stack/SC connection?

Am I barking up the wrong tree here? If so, would the virtual appliance (on the backup host) using the 'backup' datastore as a target give me a better result?


Thanks for any comments/thoughts


Max

[Gostev]

If I restore a VM from the 'backups' vmfs to the 'Datastore1' vmfs (is this possible on the same host?) - This will effectively be LAN free?

Yes, in case with "fat" ESX restores are effectively LAN-free in this scenario (I assume both 'backups' vmfs and 'Datastore1' vmfs are connected to ESX).

Would all data be pushed through the (local) SC connection (still technically LAN free but possible big performance hit)?

Correct, all data is pushed through the (local) SC connection. There should not be performance hit though.

If I restored a VM from the 'backups' vmfs to the 'Datastore1' vmfs but specify a different host in the same cluster during this operation, will this use the network stack/SC connection?

Correct. Network stack will be used in all cases.

[maxbbs]

Ok, thanks very much

[jeromyhensley]

[merged with existing discussion]

I have a physical server that I have Veeam BR installed on. The server has two NICs, one for the corporate LAN and one for iSCSI to communicate with our production SAN. Now I have my backup jobs configured to use the iSCSI NIC for performing backups.

One thing I noticed when performing a full VM restore is that Veeam BR was using the corporate LAN nic for sending out data during the full VM restore.

Why was Veeam BR utitlizing the NIC that the corporate LAN uses instead of the NIC that's used for iSCSI?.....Heh I think I just figured out my answer.

Since the server I'm using is server 2008 and utilizing the microsoft iSCSI initiator the server has read-only access to the SAN. Which is why it's utilizing the corporate LAN for the restore and not the iSCSI LAN. Can anyone verify this?

[Vitaliy S.]

Since the server I'm using is server 2008 and utilizing the microsoft iSCSI initiator the server has read-only access to the SAN. Which is why it's utilizing the corporate LAN for the restore and not the iSCSI LAN. Can anyone verify this?

Jeromy, giving write permission to your SAN LUN wouldn't help here, as all restores are performed via host network stack right now.

[mpozar]

When will Veeam provide restores via FC? It seems that vRanger 5 now provides this capability for restores. We are a Veeam reseller in Australia and we've had a few potential customers now ask about this capability/

Have FUN!

Michael Pozar

[Vitaliy S.]

Michael, we are investigating a possibility of adding such functionality to the product, but, unfortunately, right now there are some concerns related to vStorage API architecture. For more information please have a look at this post: Two things I hoped was to be in VB&R 5

[jeromyhensley]

Vitaliy, thanks for the response.

[m.novelli]

[merged into existing discussion]

Any timeframe for direct restore to SAN?
Yesterday I had to restore a 600 GB VM to ESXi 4.1 for DR testing purpouses
Full backup took 1.45h , restore took 20 hours

Marco

[slidek9]

[merged into existing discussion]

Hiya,

Sorry if this is asked a lot.

Is it possible to do a VM guest restore via SAN with B+R v5 (as I can't find the option)?

Thanks!

[Gostev]

Yesterday I had to restore a 600 GB VM to ESXi 4.1 for DR testing purpouses
Full backup took 1.45h , restore took 20 hours

You can use Instant VM Recovery to restore VM in less than 1 min... available in every Veeam Backup edition.

[KiwiJJ]

[merged into existing discussion]

Hi,
For our disaster recovery we purchased a new Host and a new workstation.
I have installed Veeam Backup 5.0.2 onto the workstation which runs Windows 7.
The host is running ESXi 4.1 Update 1
The backup files are stored on a USB Hard Drive
When testing full VM restores I am getting a maximum restore speed of 2 MB/S. At this rate it would take days to restore our 2 TB's of data / VM's

Any idea why this would be so slow ?

Any ideas on how I can get our systems back up faster in a disaster recovery situation ?
One scenario I can think of would be to install vCenter Server on the workstation.
Use Instant VM Recovery and then use vCenter to move the datastores.
How would this work running a bunch of VM's off the USB hard drive, I assume they would be very slow ?

thanks,

John

[simeonh]

Restoring via the Network is such a ridiculous idea! it so slow.
i'm sure the restores in the previous versions of veeam were much faster. has this limitation only been added for Veeam v5? If it doesn't get fixed we may have to look into other backup software, because restoring quickly is pretty critical!

[Vitaliy S.]

Simeon, no changes were introduced to the restore engine, it is still the same. Try to upload any file with vSphere Client - Datastore Browser and you'll see what speeds you should expect with your hardware and hosts version. Please note that restoring to ESX host with service console credentials specified will make the restore quicker.

As for direct SAN restores, please look through this post. On top of that, have you considered using Instant VM Recovery to get your production back to the running state?

[Vitaliy S.]

When testing full VM restores I am getting a maximum restore speed of 2 MB/S. At this rate it would take days to restore our 2 TB's of data / VM's

Any idea why this would be so slow ?

Could you please tell me what performance rates do you get while uploading any file from the same USB drive via vSphere Client -> Datastore Browser?

Any ideas on how I can get our systems back up faster in a disaster recovery situation ?
One scenario I can think of would be to install vCenter Server on the workstation.
Use Instant VM Recovery and then use vCenter to move the datastores.

Yes, in order to bring all VMs back to a production environment much faster, I would suggest using Instant VM Recovery, then use either replication job or storage VMotion to relocate those VMs.

How would this work running a bunch of VM's off the USB hard drive, I assume they would be very slow?

Yes, it will be slow, however the performance heavily depends on the read/write rates of your USB hard drive.

[KiwiJJ]

Hi Vitaliy,
I found the problem (I thought I was connected to a 1 GB switch but it was a 100MB one) and am now getting restore speeds between 50-150 MB/s.

thanks,

John

[Gostev]

This is by far most common cause for performance issues. In your case this was easy to spot, however what often happens in other cases is ports failing over from 1Gb to 100Mb for good (cabling issues) or no good (some weird port defect) reason. Glad you managed to nail this down reasonably fast. Thanks for taking time to update this thread with the resolution.

[ca107]

@Marco

were you able to improve your restore speed?

[m.novelli]

@Marco

were you able to improve your restore speed?

Not using traditional Veeam Restore metod, it's still via network and so slow (but this is an ESXi issue).

Using Instant Recovery the DR test is lightning fast, but then you need to Storage vMotion the VM to production infrastructure

I hope for FC and iSCSI direct restore for next major release

Marco

[Gostev]

I hope for FC and iSCSI direct restore for next major release

See link to the corresponding discussion topic just a few posts above.