running 1 primary replication appliance and running 1 add-on appliance

[piedthepiper]

Hi Guys,

If I was to run one main appliance which also contains the VRMS server and then add a add-on appliance to help load balance....my question is this

If I did this at the source site and the target site, if I was to lose the main appliance due to say the datastore it was on died for example, would the addon take over full replication duties?

So

VRA1 (datastore1)+ VRA-addon1(datastore2) running at site A

VRA2 (datastore3)+ VRA-addon2(datastore4) running at Site B

If I was to lose either VRA1 or/and VRA2 because their datastores died, would all the replications that were managed by them be picked up by the addon appliances (VRA-addon1/2)?

[admin]

Hi,

Unfortunately there is no automatic failover of replications between VRS. I'm suggesting you instead of this to configure your replications with offline copies so in case of some disaster (you lost VRA or VRS) you will be able to quickly deal with this case and even if you re-setup the whole VR setup (worst case) you will not transfer the whole disks across the network.

Thanks,

Dzhem

[piedthepiper]

Hi,

How do I configure offline copies? My understanding of replication is that it replicated the initial VMDKs to the target site, and then syncs changes as required. So it I was to lose the source datastore/VRA at the soruce site, the VM could be turned on at the target site and be up and running again simply minus whatever changes didn't get chance to sync because of the RPO/RTO?

[vbrowncoat]

Not sure what you mean by off-line copies. Regarding availability of the VR appliances check out: vSphere Replication Appliance Failure Prevention and Recovery | VMware vSphere Blog - VMware Blogs

For general questions on VR check out the FAQ here: vSphere Replication Frequently Asked Questions (FAQ)

If you still have questions post them here

[piedthepiper]

I was just going on what the other person said.  That first article is super interesting.

I encountered something similar in my test environment.  My iscsi storage went down and took one appliance with it,  the only way to get it working again was to unregister it from vcenter clear the built in database and resync all over again and restart the replications.

Also does anyone know how losing the vrms is impacted when using srm with vSphere replication.  Would you still be able to force the recovery at the target site, if you lost the vrms at the source?

[mvalkanov]

Hi,

If VRMS and vCenter at the recovery site are up - you can perform the recovery. VRMS at the source site is not relevant during recovery.

If VRMS and/or vCenter at the recovery site are down - you still have the replica base disks and some redo logs on top of them. The content at the base disks might not be consistent and manual recovery (without VRMS and vCenter), although somewhat possible, is error prone and not officially supported.

Regards,

Martin

[mvalkanov]

About the offline copies:

When you configure replication, you can manually place a copy of the vmdks at the target datastores and instruct VR to use them as seeds (offline copies). When you stop (unconfigure) the replication, if seeds (offline copies) have been used, VR will leave the replica base disks at the target datastore. If seeds (offline copies) have not been used, VR will remove all replica files, including the base disks. Please note that if you remove the source VM from vCenter inventory (even if the source host shows as disconnected), before initiating disaster recovery, and source site VRMS is up, VR will automatically stop (unconfigure) the replication and proceed with the cleanup as described above.

[piedthepiper]

Hi Martin,

So in the article listed above

With the VRMS appliance offline, it is not possible to recover VMs with vSphere Replication. There are some third party articles out there that suggest it is possible to manually alter the files at the target location that make up the replica and recover a VM. While this is possible, I do not recommend relying on that recovery method as it is a very manual process and, more importantly, not supported by VMware.

This is different to what you have stated? Is this article going into detail about the source VRMS and not the target site VRMS?

[mvalkanov · Accepted solution]

Hi,

It is not different from what I described.

Recovery is driven by VRMS at the target site.

If target site VRMS is down - operation can not be invoked from UI, there is no one to tell VR server to consolidate the redologs on top of the replica base disks and no one to register the resulting .vmx and .vmdks as VM in vCenter inventory.

Regards,

Martin

[piedthepiper]

Thank you for clearing it up for me, i was getting a bit puzzled haha:smileylaugh: