Does anyone have suggestions on how to best implement a VMware DR plan leveraging Nimble replication but without SRM?

[lizard0f0z71]

I have a client that needs to develop a DR playbook and I am trying to help simplify it. SRM is not an option. They have set up separate volumes for each VM because they used to use SAN replication and keep the replicated volumes attached to the DR vCenter, but Nimble does not allow replicas to be mounted so their DR plan just became more complicated. They now have to trigger the fail over and then configure the DR Nimble array connectivity for each volume and rescan in vCenter before they can begin where their existing DR Playbook currently starts. Of course this can be scripted, but I would prefer a more client maintainable option.

One alternative we have is to reconfigure Veeam to do the replication and DR prep. It has advantages and disadvantages, but does not fall into the original scope of this project.

Any suggestions are welcome.

Bill

[jtmoree23]

We are in a similar situation as we work on our own DR planning.  I have discussed with my boss that without SRM it's going to take a lot of manual work and he is OK with that.  Our plan is on failover to manually mount all of the volumes on the DR vcenter setup then manually add and turn on every critical system.  We'll have to keep lists of servers with boot order.

Any other suggestions would be welcome also by us.

[ccolht99]

I'm doing this with Powershell. It's not trivial but very doable. I originally wrote the script for NetApp but we've switch to Nimble so I recast it for that. We replicate the volumes to the DR site.

The script

1. finds the appropriate volumes
2. clones one of the snapshots (e.g. latest, oldest) if only testing
3. adds the new volumes to the init group
   1. Here's where it gets tricky. The customer we're working with uses one vol per vm. There are 40+ volumes. If you present that many volumes to esx and then scan for datastores, the process times out. So I add them 10 at a time.
4. rescan the storage adapters to see the volumes
   
5. attach the datastores and rename them to the volume names (optional but easier on the humans)
6. add the datastores to a storage cluster (makes cleanup easier, assuming you have the license)
   
7. search the datastores for vms and adds them to inventory
8. fixes the port groups as they don't match between sites (found that I could not alter the port group from powercli while a vm was in a vapp)
9. puts all the vms into a vApp and then starts it (this assumes that start order doesn't matter. YMMV)
10. answers the Move/Copy question on each one with Move
    

The script, in the Nimble form, has only been run during DR tests with volume clones. I run it in PowerGUI so I can manage it, setting break points to validate operations as it runs. I am somewhat of a hack in PowerShell and stood on many shoulders to get here.

https://www.dropbox.com/s/ax3fe68m3a40ilu/nimbledrvms.v0.9.ps1?dl=0

This takes about 2 hours to run in our environment. The NetApp version with 6 NFS volumes took about 30 min. Most of the time is lost in the volume discovery process.

Purists please be gentle.

[bgrieve65 · Accepted solution]

Chuck's solution, or something very similar, leveraging scrips is the most common method of creating DR workflows with Nimble without the help of 3rd-party software. While there are several good solutions on the market, many of my customers tend to like Zerto these days.

[ccolht99]

Thanks for pointing out Zerto. Hadn't heard of it. We're on mostly Nimble today but maybe not tomorrow. Being able to (efficiently) work across platforms is always preferred. With DR it is especially important to have flexibility. That's why I generally recommend file based backups for archiving rather than array based. Take advantage of the storage system for fast backup and restore but always have a copy of the data that is storage agnostic, even if it isn't as current.

[jtmoree23]

I would prefer to keep all backup data on slow (low cost) storage but in the case of DR actually happening we need to be able to run the entire infrastructure from the DR site.  Are you using Nimble at the DR site?

[ccolht99]

Yes, in this case we are using Nimble for DR. We are taking advantage of array replication for specific vms. The customer has their own archiving system for long term storage. In general, we use a combination of array and file to tape based backups. The array backup will get our critical systems back online as quickly as possible. The tapes provide less expensive archiving with better version and retention control for data that requires it. You don't want to maintain tons of point in time copies on live storage.. Nimble is weak in this area. Something like NDMP would be nice. We use <insert major multi-platform file  backup system> for offsite archiving. Been through 4 storage platforms over the last decade and we can still recover all the data.

[lizard0f0z71]

Backups are not on the Nimble, but the DR site is using Nimble and full hourly replication is in place. I am primarily trying to simplify their DR playbook. I can script it, but I prefer not to leave an unsupported script behind. I will probably write the script and provide the manual process as a backup.

Bill Baker

[SprinkleJames]

Chuck,

Thanks for your post. I'm interested in seeing your script - would you be willing to re-post it? Your link no longer works.

[Valdereth]

If you're using Veeam keep an eye out for Nimble integration with 9.5, this is going to provide some great options for Nimble/Veeam customers:  Veeam integration with Nimble Storage