AnsweredAssumed Answered

GAD global active-device: failback scenario?

Question asked by Steffen Zimmermann on Jan 12, 2016
Latest reply on Jan 13, 2016 by Dang Luong

Like • Show 1 Like1
Comment • 10

Hello Community!

Just testing GAD these days with different failure scenarios.

Wondered, how to failback, if, e.g., the "primary site" P-VOL is PSUE and the S-VOL is SSWS.

Machines are up and running from the "S-VOL"-site.

The User Guide just said "resynchronize pairs".

I've done it with -swaps from the S-Vol Site (-> now the new P-Vol), after resync, the status is PAIR again -> P-Vol now on the "secondary system".

The same to bring back the original P-Vol to the P-Vol state.

ESX-Hosts are fine, all paths are online again. Unfortunately, Hyper-V-Hosts (WITHOUT HDLM-Driver) crashed with the above method.

Could you help, what is wrong?

Thank you.

kind regards

maoin

No one else has this question

Outcomes

10 Replies

Dang Luong Jan 12, 2016 10:03 AM
Mark Correct
Correct Answer
Dear Maoin,

By crashed, did you mean the Hyper-V hosts rebooted abruptly? By any chance, did you remove paths to the P-VOL before performing the recovery procedure?

Regards,
Dang Luong
Like • Show 0 Likes0
Actions
Steffen Zimmermann Jan 12, 2016 10:14 AM
Mark Correct
Correct Answer
Hello Dang Luong,

no, I have not removed any paths. And ESX hosts survived this procedure sucesssfully.
And yes, the Hyper-V host lost connection the LUNs.
Generally, is this the right procedure to recover from an failure with GAD pairs?

thank you very much
maoin
Like • Show 0 Likes0
Actions
- Dang Luong @ Steffen Zimmermann on Jan 12, 2016 10:32 AM
  Mark Correct
  Correct Answer
  Hi Maoin,
  
  There are quite a few possible failure scenarios, e.g. primary path failure, secondary path failure, replication link failure, quorum failure, primary storage failure, etc... I assume you're testing primary storage failure, in which case the complete recovery procedure is:
  
  When the primary storage system is powered off, delete an alternate path (logical path) to the P-VOL, and then turn on the power.
  Using the alternate path software, distinguish the volumes which are not able to be accessed from the server.
  Using the alternate path software, delete the alternate paths to the PVOL. If you cannot delete the alternate paths, detach all channel port connection paths (physical paths) which are connected to the server at the primary site.
  Turn on the primary storage system.
  Recover the primary storage system. For details, contact Hitachi Data Systems customer support.
  Recover the physical path between the primary storage system and the secondary storage system.
  If S-VOL pair status is PAIR, suspend the pair specifying the S-VOL.
  (VSP G1000 only) Confirm the quorum disk status. If the quorum disk is blocked, recover from the blockade.
  Resynchronize or re-create a pair using the procedure in the following table whose pair status and I/O mode match your pair's status and I/O mode.
  If the alternate path to the P-VOL has been deleted, add the alternate path.
  If you have detached the channel port connection paths of the primary site, restore all channel port connection paths to their original status, and then add the alternate path.
  Using the alternate path software, add the alternate path deleted at step 1 to the P-VOL.
  
  Regards,
  Dang Luong
  1 person found this helpful
  Like • Show 0 Likes0
  Actions
  - Steffen Zimmermann @ Dang Luong on Jan 12, 2016 11:17 AM
    Mark Correct
    Correct Answer
    Hello Dang Luong,
    
    thank you. What we have done:
    
    Blocking all Remote links for replication. After that, the P-vol was PSUE and the S-Vol SSWS. The host lost the two paths to the primary system, but was up and running.
    
    To recover, I released the blocked Remote links and after that:
    
    pairresync -IHxx -g Test -d vol1 -swaps
    -> resync was done, status PAiR again
    pairresync -IHyyy -g Test -d vol1 -swaps
    -> to bring the p-vol to the primary site again
    
    MAybe I was to fast with these commands?
    
    IHxx -> secondary system
    IHyyy -> primary system
    
    kind regards
    maoin
    Like • Show 0 Likes0
    Actions
    - Dang Luong @ Steffen Zimmermann on Jan 12, 2016 12:02 PM
      Mark Correct
      Correct Answer
      Hi Maoin,
      
      Oh then my assumption was incorrect. This failure scenario would be path failure from secondary to primary storage system, AKA replication link failure.
      
      Your commands are correct. One thing you could do is after the first pairresync, verify that the hosts are utilizing paths to the new S-VOL (previously P-VOL). Also, HDLM's default failback time is 60 seconds so it might not have restored the original paths yet before the second pairresync was executed. This would have left the Hyper-V servers without any online paths, causing them to crash.
      
      Regards,
      Dang Luong
      1 person found this helpful
      Like • Show 0 Likes0
      Actions
      - Steffen Zimmermann @ Dang Luong on Jan 12, 2016 12:36 PM
        Mark Correct
        Correct Answer
        Again, thank you. I will try that tomorrow.
        One more question: is HDLM on the host necessary? We tested with standard drivers.
        
        regards
        maoin
        Like • Show 0 Likes0
        Actions
        Dang Luong @ Steffen Zimmermann on Jan 12, 2016 5:04 PM
        Mark Correct
        Correct Answer
        Hi Maoin - Are you using cross paths? What is your microcode version?
        
        Regards,
        Dang Luong
        Like • Show 0 Likes0
        Actions
Steffen Zimmermann Jan 12, 2016 10:36 PM
Mark Correct
Correct Answer
Hi Dang Luong,

yes, the single host in this scenario is connected to the primary and secondary system with two paths each. Microcode version: 83-01-24-20/00

thank you
maoin
Like • Show 0 Likes0
Actions
- Dang Luong @ Steffen Zimmermann on Jan 13, 2016 9:40 AM
  Mark Correct
  Correct Answer
  Hi Maoin,
  
  You'll need HDLM to avoid sending I/O to the cross paths, at least for now. Support for ALUA with GAD on the VSP Gx00 platform is not out yet.
  
  Regards,
  Dang Luong
  1 person found this helpful
  Like • Show 0 Likes0
  Actions
Steffen Zimmermann Jan 13, 2016 2:16 AM
Mark Correct
Correct Answer
Just a short update:

As Dang Luong said, after the first pairresync -swaps we started a Rescan on the Hyper-V Host. Now, all 4 paths were visible. The second pairresync -swaps (to bring the P-Vol back to the primary site) was after that successfully, too.

Thank you for your help.

Kind regards
maoin
1 person found this helpful
Like • Show 1 Like1
Actions

Jive

Hitachi Vantara Community

GAD global active-device: failback scenario?

Outcomes

Related Content

Recommended Content