To suspend and resume snapmirror in the middle of initialization

[MARIKOTAGAWA]

Hi,

I asked about snapmirror behavior with opion "abort" and "initialize".

Accoring to YIshikawa, snapmirror is to resume from the last checkpoint instead of from the scratch.

http://community.netapp.com/t5/Data-ONTAP-Discussions/To-suspend-snapmirror-in-the-middle-of-initialization/m-p/101801#M20626

I had a snapmirror initializing by "snapmirror initialize -S <src> <dest>"

Then I aborted the initialization by "snapmirror abort <dest>"

Then resumed by "snapmirror initialize -S <src> <dest>"

What happend was the snapmirror initilization started from the scratch. It didn't work as YIshikawa explained.

Was my operation wrong? Please find attached logs.

Can anybody instruct me how to suspend snapmirror initialization and restart from the checkpoint?

Here are logs.

---------------------------------------------------------------------------------------------------------------------------------

Before aborting:

    <source filer>:dir_17  <dest filer>:dir_17  Uninitialized  516:54:09  Transferring  (1811 GB done)

    #snapmirror  abort dir_17

When the snapmirror aborted.

     <source filer>:dir_17  <dest filer>:dir_17  Uninitialized  516:55:20  Idle with restart checkpoint (at 1810 GB)

Then

     #"snapmirror initialize -S <source filer>:dir_17  <dest filer>:dir_17

When I check after 5 hours

    <source filer>:dir_17  <dest filer>:dir_17  Uninitialized  519:58:05  Transferring  (28 GB done)

Currently

     <source filer>:dir_17  <dest filer>:dir_17   Uninitialized  565:27:18  Transferring  (662 GB done)

---------------------------------------------------------------------------------------------------------------------------------

Thank you in advance.

Mariko

[mbeattie]

Hi Mariko,

It might be related to this:

https://kb.netapp.com/support/index?page=content&id=2012288

Bug ID: 83700

Description
    
 If you abort a "snapmirror initialize", the snapmirror will have the state
 "idle with restart checkpoint (at XX mb)".  You can then restart the
 SnapMirror from where you left off with a "snapmirror initialize" again.
 However if a scheduled update occurs while the initialization is idle,
 the state will be "pending with restart checkpoint (at XX mb)".  If you
 do a "snapmirror initialize" now, it will restart the entire transfer
 again from the beginning.

Workaround
    
 Before attempting to restart "snapmirror initialize", execute:
 
 filer> snapmirror off
 filer> snapmirror on
 
 The state will revert (from "pending...") to "Idle" and the "snapmirror
 initialize" transfer may be restarted at the checkpoint rather than
 starting over again from 0.

/matt

[MARIKOTAGAWA]

Hi, mbeattie.

Thanks so much for your reply and the good article related to my problem.

-------------------------------------------------------------------------
Cause
    1. BUG 83700
    2. The SnapMirror relationship being initialized is the last leg in the cascading chain and the storage system in the middle has performed the scheduled updates while the SnapMirror was in the 'idle with restart checkpoint' state.
    3. SnapMirror source controller name changed. There is no workaround for this case. SnapMirror initialization has to start from scratch.

-------------------------------------------------------------------------

Actually, my case is the second one. I've tried to initialize the last leg of the cascading snapmirror; c:vol1.

a:vol1 -> b:vol1 -> c:vol1

As until completing the snapmirror initialization of c:vol1, the middle of snapmirror a:vol1 -> b:vol1 becomes pending.

Due to business requirement, b:vol1 should have the copy of a:vol1. Their difference is only allow for 1day.

Therefore I have to abort b:vol1 -> c:vol1 to complete a:vol1 -> b:vol1 everyday.

I read the Solution for 2.

https://kb.netapp.com/support/index?page=content&id=2012288

It's a bit completed but no option. I'm going to start snapmirror initialization of c:vol1 from a:vol1.

Then wait the completion (it will take about 10 days) and change the snapmirror source from a to b. I hope this solution works without error.

I'll let this community knows when I solved the issue.

Thank you mbeattie and community member.

Mariko

[widelinks]

Hi All,

Thanks for sharing the post.

I have observed same behavior under Data OnTAP 8.2.3P6 while initilializing snapmirror remote replication.

We were able to re-initialize the interrupted snapmirror initial synchronization successfully using the workaround mentioned under BUG 83700.

Thanks

Sivaramakrishna.