BHORAN1
2 Iron

failing over individual devices in SRDF/A session

Jump to solution
I'm a bit unclear about what flexibility is possible in SRDF/A.

We have a RDF group that is setup with async. There are a number of devices all running async with consistency enabled. We would like failover a select few devices for a DR test.

Is this possible? Or do all the devices in the SRDF/A session need to be managed as a whole??
Tags (2)
1 Solution

Accepted Solutions
MikeMac1
3 Argentium

Re: failing over individual devices in SRDF/A session

Jump to solution
If you wanted to create a second async group and wanted to move the selected devices over to it you would have to bring down your SRDF/A session in order to do it. So in essence your procedure would be to stop SRDF/A, make your configuration changes to move the devices over to a new RDF group, then you could resume your SRDF/A session then work your failover devices using the new group you just created.
7 Replies
MikeMac1
3 Argentium

Re: failing over individual devices in SRDF/A session

Jump to solution
The SRDF/A sessions need to be managed as a whole. You would have to take the session out of SRDF/A to be able to failover the select few devices for your DR test.
BHORAN1
2 Iron

Re: failing over individual devices in SRDF/A session

Jump to solution
Okay, so I could set the group to sync and failover the individual devices.

If I wanted to keep them all async could I create a new RDF group that is async and move the individual devices into this RDF group and fail them over?
MikeMac1
3 Argentium

Re: failing over individual devices in SRDF/A session

Jump to solution
If you wanted to create a second async group and wanted to move the selected devices over to it you would have to bring down your SRDF/A session in order to do it. So in essence your procedure would be to stop SRDF/A, make your configuration changes to move the devices over to a new RDF group, then you could resume your SRDF/A session then work your failover devices using the new group you just created.
BHORAN1
2 Iron

Re: failing over individual devices in SRDF/A session

Jump to solution
Okay so moving the individual devices to new group would involve:

1) Create new RDF group and set as acp_disk
2) Suspend all devices on my exsting async RDF group
3) Movepairs from the existing async RDF group into the new RDF group
4) Return existing async group to established/consistent
5) Establish devices in the new async group
6) Failover devices in new async group as desired
7) Reverse process to move devices back into the main async RDF group
MikeMac1
3 Argentium

Re: failing over individual devices in SRDF/A session

Jump to solution
That pretty much sounds right. I'm not a script expert so someone else on the forum's might chime in if we are missing any steps.
0 Kudos
xe2sdc
4 Ruthenium

Re: failing over individual devices in SRDF/A session

Jump to solution
Bhoran you are almost right .. some tiny improvements...

1) create new RDF group
2) put all devices on existing async RDF group in acp_disk mode
3) suspend devices you want to move
4) movepairs (if you have code 5773 or better) on new RDF group
5) put all remaining devices on "old" RDF group in old mode
6) establish devices in new RDF group (they will still be acp_disk)
7) put devices in new RDF group in async mode

Now you have a brand new RDF group you can manage as a whole thing independently from the old one.
If you want to merge them back:

1) put all devices in both old and new RDF group in acp_disk
2) suspend all devices in new RDF group
3) movepairs (if you have code 5773 or better) from new to old RDF group
4) establish moved devices in old RDF group
5) put all devices in async
6) delete unneeded RDF group :-)
Siaotong
1 Copper

Re: failing over individual devices in SRDF/A session

Jump to solution

If you're working on a V-Max with 5874, then you can use the -cons_exempt flag as below (excerpt from SE SRDF Family CLI Product Guide v7.0)

SRDF/A and the consistency exempt option

When an RDF group is supporting an active SRDF/A session, control operations

must be targeted at all device pairs in the session, or the session must first be made

inactive by suspending the links between all device pairs in the session before

attempting to control a subset of the device pairs.

The consistency exempt feature, available with Enginuity 5874, relaxes that

restriction. Devices marked consistency exempt can, in some cases, be controlled

independently of other devices in the active SRDF/A session. The

-cons_exempt

option flags devices targeted by the command as consistency exempt. Enginuity

automatically clears the consistency exempt status when the affected device pairs

have become consistent and two cycle switches have subsequently occurred.