Comprehensive data protection for all workloads
Post Reply
nickf
Influencer
Posts: 19
Liked: never
Joined: Jun 22, 2010 3:11 pm
Full Name: Nick Ferrar
Contact:

Manually bringing up replicas

Post by nickf » Sep 08, 2010 8:58 am

Hi,

I'm in the process of setting up Veeam 4.1 to replicate several VMs from our production site to our DR site however in the case of a real DR I don't see the "Perform Failover" function as an option (the Veeam server runs in the site that's just burnt down etc...). However I can't see any procedure documented (in the User or Eval guides) about how to correctly do a manual failover. As a test we just powered up a replica VM without a NIC connected (using the vCenter in the production site) and when prompted selected that it was a copied VM. The VM started fine and looked OK and once powered off I started the replication job again and it seemed to work fine and only did (as hoped) incremental changes, but from what I can see in the User guide if you bring the VM up manually you then break the replication ability and have to re-seed a new replica (or am I reading it wrong)?

Fortunately our VMs are pretty static so our DR test plan doesn't need to include a failback so it *should* be a fairly simple test but I want to make sure we're using the correct procedure when manually bringing up the replica VMs at the DR site and if we were just lucky or if it's been fixed now so if you do bring replicas up manually you can continue with incremental replication once the replica is powered down and the link to the production site is restored.

Also is there anything I need to be aware of in order to replicate our vCenter VM (I noticed in a forum post selecting the ESX host rather than vCenter VM when backing it up but does that apply to replication to? we're only using Veeam for replication in this environment as it's a mix of physical and virtual servers so we're using NetBackup for all backups for consistency)

One other thing - as you have to select the target ESX host for replication I presume if that host is shutdown (for patching etc.) then the replication job will fail, is it possible at that point to modify the replication job to point to another ESX host at the DR site (datastore would stay the same) and for replication to continue where it left off or does it need to do a full seeding to a new replica again?

Thanks,

Nick

Vitaliy S.
Product Manager
Posts: 22998
Liked: 1556 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: Manually bringing up replicas

Post by Vitaliy S. » Sep 08, 2010 10:12 am

Nick,

1. Please have a look at the procedure of testing replicated VM without using Veeam backup console:
http://www.veeam.com/forums/viewtopic.p ... ing#p10941

2. Yes, this also refers to replication jobs as well.

3. Yes, you may change the destination and in the replication job advanced settings not to brake the replication cycle.

Thank you!

nickf
Influencer
Posts: 19
Liked: never
Joined: Jun 22, 2010 3:11 pm
Full Name: Nick Ferrar
Contact:

Re: Manually bringing up replicas

Post by nickf » Sep 08, 2010 11:12 am

Thanks for the clarification!

I did have another question... in order for the dedupe to work should the replication job include all the VMs? I was concerned that if one of the VMs in a replication job failed then the whole job would fail so I created 20 different replication jobs each with one VM in. Is that the best way to do it or should all the VMs be in one job?

Alexey D.

Re: Manually bringing up replicas

Post by Alexey D. » Sep 08, 2010 12:13 pm

Hello Nick,

For best dedupe and compression ratios you should place all VMs inside one job (this is not designed to work across multiple jobs). In case of replication, only rollbacks (VRBs) are affected by deduplication and compression.

Concerning your second question, there is the "retry failed VMs" option on the scheduling tab of job's properties - you may want to set this number to 3 or 4, which should allow you to have an overall success status for your job. Hope this helps!

Gostev
SVP, Product Management
Posts: 24797
Liked: 3557 times
Joined: Jan 01, 2006 1:01 am
Location: Baar, Switzerland
Contact:

Re: Manually bringing up replicas

Post by Gostev » Sep 08, 2010 12:29 pm

nickf wrote:I did have another question... in order for the dedupe to work should the replication job include all the VMs? I was concerned that if one of the VMs in a replication job failed then the whole job would fail so I created 20 different replication jobs each with one VM in. Is that the best way to do it or should all the VMs be in one job?
This is fine to do since dedupe does not apply to replicas.

nickf
Influencer
Posts: 19
Liked: never
Joined: Jun 22, 2010 3:11 pm
Full Name: Nick Ferrar
Contact:

Re: Manually bringing up replicas

Post by nickf » Sep 16, 2010 2:06 pm

Just thought I'd say an internal DR test worked fine (recovered and tested 16 VMs in 1.5 hours, certainly beats tape restore which would involve 1.5 hours just to get the tapes back on-site).

The only thing that confused me was mopping up post-test, I reverted the snapshots back to the pre-test ones but looking at them via vCentre snapshot manager it appears you're still below the snapshot point (it doesn't move the 'you are here' flag) but in the end I just deleted the pre-snapshot anyway to clean things up ready for the next time. The replication has worked to the VMs since so I assume that was right thing to do...

I have to say I'm also surprised at just how much the Widows 2008 R2 VMs are changing, I figured it would be around 10MB of changes per hour on a VM not in use but we're seeing a minimum of 50MB and things like a Forefront gateway server is changing around 200MB an hour (presumably mostly from logs). This means our 10Mb link is a bit undersized so I've been playing around extending the windows to 4 hours and trying to stagger them but as I've also put in an overnight window where the replication is disabled (to avoid conflicts with our NetBackup jobs) it means they all try and start at 2am when the window reopens - not good when 16 different jobs try and start ;) I'm going to wait and see if scheduling is improved at all in v5 otherwise we'll switch to using Control-M and Powershell.

meier
Influencer
Posts: 10
Liked: never
Joined: Dec 21, 2010 9:53 am
Full Name: Holger Meier
Contact:

Starting replicated virtual machines without VEEAM

Post by meier » Feb 22, 2011 3:20 pm

[Merged into existing discussion]

Is it possible to manually start a replicated virtual machine in the vCenter or do I use the restore wizard of VEEAM?

Bunce
Expert
Posts: 259
Liked: 8 times
Joined: Sep 18, 2009 9:56 am
Full Name: Andrew
Location: Adelaide, Australia
Contact:

Re: Manually bringing up replicas

Post by Bunce » Feb 22, 2011 3:27 pm

As well as the info above, this is also noted in the FAQ under REPLICATION (stickied at top of forum)

Post Reply

Who is online

Users browsing this forum: Bing [Bot] and 52 guests