Testing CA Failover

 
Abdul Rashid
Frequent Advisor

Testing CA Failover

We have 2 EVA8100 with XCS V6220 and configured with Continouos Access. I am trying to fail over one of the DR groups from the source EVA8100 to the destination EVA8100 for testing purpose. All the DR groups are showing as normal and healthy. When failing over I have chosen fail over and resume replication after failover and get a warning message saying all the data on my original source EVA8100 will be over written. I do not want to lose my original data.

My question is, will I lose all my orininal data, I am a bit confused with the message I am getting.

Thanks
8 REPLIES 8
Del_3
Trusted Contributor

Re: Testing CA Failover

Abdul, Sometimes a failover forces a full copy. This would overwrite your original source. Are you in Async or Sync mode? You must be able to transition Async to Sync to prevent a full copy durning a failover test.

How about making a Snapclone of the original source before you failover? That will preserve your data.

Can you provide us with the complete message.

Dileep..
Trusted Contributor

Re: Testing CA Failover

Hello Abdul Rashid,

Below is the procedure i would suggest for a planned failover,

1) Shutdown all the host at source (hope you are not using any metro cluster SW)
2) Change replication to Sync
3) Wai till all the data in cache / log disk flush to destination.
4) Ensure that LOG disks at source are empty.
5) Do the failover
6) start the host at destination.

I am quite sure that this process wont start full copy in either way.

Doing a failover before flushing the data to destination EVA is not suggested.

Regards
Dileep
Abdul Rashid
Frequent Advisor

Re: Testing CA Failover

Thanks for both of your replies. To answer your questions, all the DR groups are configured in sync mode and we still get a warning message saying all the data in the original source will be over wriiten, (which I do not want it to happen). Some of the luns in the DR groups are of maximum size of 2 TB and we do not have enough space to create snap clones for all the luns.

The method I followed to initiate failover was.

1) Shutdown the server that owned the luns in the DR group.
2) Made sure there was no data in the logs and all the members within the DR group were normal
3) From the destination EVA8100 initiated the failover with continue replication after failover.

And at this stage we are getting the warning that all the data will be overwritten in the source EVA.

Hope this helps.
KurtG
Regular Advisor

Re: Testing CA Failover

Are you (100%) sure that the DR luns are not currently executing av full copy? (Check the DR groups - Members Full copying = 0 on source array)

KurtG
Abdul Rashid
Frequent Advisor

Re: Testing CA Failover

Yes 100% sure that the state all the members within the DR group in normal and the progress showing the status of n/a
KurtG
Regular Advisor

Re: Testing CA Failover

What version of Command View EVA are you running?

I do no see any reson for this message.

KurtG

BTW! You should consider you points assignments policy. Assigning points to 5 of 45 questions really does not help you get the answers you are seeking.
Abdul Rashid
Frequent Advisor

Re: Testing CA Failover

Command View V9.0.083. Yes you are right I was also not expecting to get this warning message and that is why I am worried to carry out the failover testing.
DogBytes
Valued Contributor

Re: Testing CA Failover

Abdul,

it makes sense if you continue replication after failover, the source would be overwritten. After failover, the original destination becomes the source and with replication resumed, the original source is the new destination. If you want to be 100% safe, expand your array so you can do snapclones of your data (ideally offline) before DR testing. Remember to take the time the snapclones will require into consideration.

Hope that helps,

- R.