Subscribe

How to planned failover of snap-mirrored volume

Hi

We currently have 3 FAS 3140 filers all running 7.3.2, a clustered pair in our external Datacentre (NAP01 / NAP02) and a standalone filer (NAP03) in the DC at our office, they are run active / active with 10Gig connectivity between DC's.

We want to completely wipe the clustered filers at the external DC and at the same time upgrade the firmware / software and join to our new domain, currently they run a number of volumes that contain CIFS / NFS shares, these are snapmirrored to the standalone filer.

Whilst these are being rebuilt i need to stop the snapmirror and make the volumes on the standalone filer active, once rebuilt we can then look at creating new snapmirror relationships to get them back to the clustered filers.

We have inherited these from a take over so i don't have much experience on Netapps and picking it up as i go, I tried on a test LUN but ended up with in a bit of a mess and the snapmirror said broken-off and i cant do anything with it, i cant risk doing that on the live data!!

Can someone please give me the steps needed to fail this over without data loss

Many Thanks

Nick

Re: How to planned failover of snap-mirrored volume

1. Stop access to source volumes from clients

2. Do final snapmirror update on destination

3. Do snapmirror break on destination

Key is to ensure no data is changed on source after final update. Any changed data will obviously be missing in destination.

Re: How to planned failover of snap-mirrored volume

Hi Aborzenkov

Thanks for the reply

Is there an easy way to stop the access of would i have to remove the cifs and NFS shares

When i break the Snapmirror will that automatically bring the volume live on the current secondary site

What happens to the volume that is currently live, does it take that offline or make any changes?

Thanks

Re: How to planned failover of snap-mirrored volume

When you break SnapMirror, source volume is not affected at all. Destination volume becomes read-write. I'm not sure I understand you question about "live" - what do you mean?

One possibility to stop access is to shutdown clients