2010-01-17 08:00 PM
We have a few vFilers that serve a few volumes, each for different businesses. Now one of the businesses want to conduct a DR test. We have in the past run a vFiler to vFiler DR failover and it worked, now worries. This was when we were serving one business only. Right now we are serving multiple businesses thus we cant do a vFiler DR test as it will impact all businesses.
I ve given this some thought and think that we should be able to run a <snapmirror break> on the volume we want to DR test, then assign that volume back to physical filer and share
it out from there. see attachment.
I ve done some testing where I have created a test volume, mirrored it to the DR vfiler which is not running. I ve broken the mirror and now I am at the stage where I want
to hand the volume back to vfiler0:
FAS> vfiler remove htrifil0004 vfiler_dr_test
Cannot remove resources from htrifil0004. Vfiler is part of the DR
configuration for a remote vfiler.
Does anybody know how to get out of this one?
Solved! SEE THE SOLUTION
2010-01-18 07:40 AM
I have done a ton DR tests and I have never seen this before.
I have seen an issue with auditing running and not allowing (something like) this, but the message was different.
One thing you may try, is rebooting the filer itself (if this is allowed) and commenting out the vol in the "rc" that way it will not start up, and you may be able to move the relationship after manually starting the volume, or vice versa.
Another thing I am thinking is a common snapshot is tying up the volume and it is either (busy) or needs to be deleted.
I am curious to what version of the OS you are running.
2010-01-18 08:15 AM
The vfiler dr feature won't allow moving a volume unless you destroy the dr vfiler. Some workarounds you can use... The easiest would be FlexClone if you have the license... flexclone the volume you want to test and the mirrors keep running to all volumes even the test volume.
The other way is brute force, but you could destroy the dr vfiler (this won't delete the data), so the mirrors will keep running to all but the test volume... then break the one mirror you want to test... when done, recreate the vfiler with "vfiler create -r" using the -r option to rebuild the vfiler only from the root volume name... then "vfiler dr resync" to put the active vfiler in dr mode again.. it is a big workaround but would accomplish what you need keeping the mirrors running to all but the one volume...but it wouldn't be fun to do this often... flexclone would be the best way to go.
2010-01-18 01:32 PM
I do think Scott is on the money here. I was thinking of using vol copy and/or flexclone on the way to work this morning however that poses another problem: the volume will be used for production and thus needs to write data back to prod. before reverting and failing back. This pretty much voids the flexclone procedure as a flexlone would not take the working snapmirror relationship into consideration.
So the only workaround is the destruction of the vFiler scenario which is a potential risk to all the businesses of course, its small but try to tell the other businesses that ;-)