Load sharing SVM root mirrors and ontap upgrades

[BenCoughtry]

Hi all, I've read that it is recommended to move LS mirror source volumes to an aggregate on a node to be upgraded last.  Would it be sufficient instead to quiesce LS mirrors, NDU and then resume relationships afterwards?  Are there dangers involved in doing it this way?

I am not worried about any temporary performance degridation by quiescing the mirrors, although existing connections must not be inturrupted as they are NFSv3 datastores for ESXi.  In fact the only reason I have LS mirrors at all on >2 node clusters is to ahear to best practices.

I'll be NDU'ing a 4-node cluster from OnTap 8.2.2 -> 8.3.2P9 -> 9.1P1.

Thanks!

-Ben C

[robinpeter]

Load-sharing mirrors for data volumes are deprecated in 9.1P1

You can no longer create new load-sharing mirrors for data volumes in 9.1P1.

Use the below command options if you want to use load-sharing relationships on data volumes and retry creating the SnapMirror relationship:

-option-name replication.ls_mirrors_on_data_volumes.enable on

Note: Existing load-sharing mirrors data volumes continue to be supported but support will be withdrawn in a future ONTAP release.

I'll recommend you remove the LS mirror and create a DR mirror for SVM-root to each node.

Here is an example:-

vserver : CIFS01
vserver root volume : CIFS01_root

node-name    Aggr-Name
cupcake-01 :: cupcake01_aggr1_6T_SATA
cupcake-02 :: cupcake02_aggr1_6T_SATA
cupcake-03 :: cupcake03_aggr1_1800G_SAS
cupcake-04 :: cupcake04_aggr1_6T_SATA

::> vol create -vserver CIFS01 -volume CIFS01_root_DPmir1 -aggregate cupcake01_aggr1_6T_SATA -size 1G -state online -type DP
::> vol create -vserver CIFS01 -volume CIFS01_root_DPmir2 -aggregate cupcake02_aggr1_6T_SATA -size 1G -state online -type DP
::> vol create -vserver CIFS01 -volume CIFS01_root_DPmir3 -aggregate cupcake03_aggr1_1800G_SAS -size 1G -state online -type DP
::> vol create -vserver CIFS01 -volume CIFS01_root_DPmir4 -aggregate cupcake04_aggr1_6T_SATA -size 1G -state online -type DP
 
::> snapmirror create -source-path CIFS01:CIFS01_root -destination-path CIFS01:CIFS01_root_DPmir1 -type DP -schedule 5min
::> snapmirror create -source-path CIFS01:CIFS01_root -destination-path CIFS01:CIFS01_root_DPmir2 -type DP -schedule 5min
::> snapmirror create -source-path CIFS01:CIFS01_root -destination-path CIFS01:CIFS01_root_DPmir3 -type DP -schedule 5min
::> snapmirror create -source-path CIFS01:CIFS01_root -destination-path CIFS01:CIFS01_root_DPmir4 -type DP -schedule 5min
 
::> snapmirror show -source-path CIFS01:CIFS01_root
 
::> snapmirror initialize -source-path CIFS01:CIFS01_root -destination-path CIFS01:CIFS01_root_DPmir1
::> snapmirror initialize -source-path CIFS01:CIFS01_root -destination-path CIFS01:CIFS01_root_DPmir2
::> snapmirror initialize -source-path CIFS01:CIFS01_root -destination-path CIFS01:CIFS01_root_DPmir3
::> snapmirror initialize -source-path CIFS01:CIFS01_root -destination-path CIFS01:CIFS01_root_DPmir4
 
::> snapmirror show -source-path CIFS01:CIFS01_root

[BenCoughtry]

As long as this would maintain alignment with netapp best practices this would be fine.  Thank you,

[mcope · Accepted solution]

Peter's response was specifically about creating Load-Sharing Mirrors for data volumes.  Unless and until you see an announcement about deprecating the use of LSMs altogether, it is still a best practice to create LSMs for SVM root volumes on every node in the cluster.  They still need to be deleted and recreated for major version upgrades of ONTAP.   Follow the instructions in the ONTAP 9 SVM Root Volume Protection Express Guide to create and maintain the LSMs.

https://library.netapp.com/ecm/ecm_download_file/ECMLP2496241 

[robinpeter]

Thanks mcope.

Even for the SVM-root volume.. you can use the DR mirror insteard of LS mirror.

in the event of a SVM-root volume recovery.. DR mirror is easier to recover.. than an LS mirror.