7 mode to cluster mode transition questions

[san456]

HI 

preparation is underway to migrate 7 mode volumes to cluser mode and have come across few questions releated to a warning reported by a precheck and cutover releated questions.  appreciate your expert opinion here. sorry for lenghy questions. appreciate your responses.

1) Is there a command to disconnect NFS client access to a 7-mode volume being migrated from the storage side during storage cutover? CIFS access to a 7-mode volume being migrated can be terminated from storage side using “cifs terminate –v vol_name”, is there a way to terminate NFS access to a volume from the storage side to prevent additional writes from being performed while final copy is in progress during storage cutover phase?

2) 7Mtt precheck reported below warning (warning no 90702)

Access to non-Unicode directories on the transitioned volumes by NFSv4 or CIFS clients might fail in 8.3.0 or later versions of Clustered Data ONTAP.

Identify the non-Unicode directories and fix the issues immediately after the transition is complete. For more information about how to handle non-Unicode directories, refer to the 7-Mode Transition Tool Release Notes.

  Is above warning ignorable. What is the criticality to address above warning for successful transition with regard to client access after the cutover?

3) I plan to use new identify for CIFS on the cluster (different CIFS server NETBIOS name in AD domain than the one 7 mode using currently) . Per 7MTT 3.3.1 user guide, new CIFS identify can be used having configured the CIFS server on the target SVMwith a new CIFS identity.  I hope not but wanted to double check, will there be any impact with regard to client access using new CIFS identify on the cluster than the one that 7 mode using after the cutover.

4) In our current 7-mode configuration, we have two controllers in HA. As part of our migration project, we plan to migrate volumes on one controller now and volume on partner controller will be migrated later in the year. We have a plan to migrate the IP address of the controller being migrated manually and configure it on cluster after 7MTT cutover task completed so clients can continue to use same IP after the storage cutover to cluster. Volumes being migrated will be taken offline as part of the 7mtt storage cutover, if no IP selected for transition, then IP on 7 mode controller will not be brought offline and it need to be brought down manually correct ?. When I bring down IP address of a 7-mode controller manually after the storage cutover, complete to reconfigure cluster LIF to use same IP, will 7-mode Ontap trigger a takeover? because negotiated failover is enabled on data interface of the 7 mode controllers, if it bring down a NFO enabled 7 mode interface , will Ontap trigger takeover of the 7 mode node where interface being brought down. (OR) taking down interface manually does not account for NFO and takeover will not be initiated.

[SpindleNinja · Accepted solution]

1) You can unmount,  if you have a host running an active session to an NFS,  I would hold whatever processes on that host is causing the IO.   

2) I've seen that one before,  I believe just a general warning.  I could not ever find a non uni code dir on a system that got that flag.  

https://kb.netapp.com/app/answers/answer_view/a_id/1029883/~/how-to-determine-if-volumes-contain-non-unicode-directories- 

3) No,  this is pretty common.  Make sure all your clients know where your new cifs is.  i.e. via changing IP mappings or updating DNS records.  

4) it shouldn't but typically i've re-ip with a new IP in that situation.   All previous CIFS shares are removed and volumes offline.    

[san456]

Thank you.

[moep]

1) You could stop nfs via command "nfs off", if unmounting on the source is no option.

[san456]

is it not necessary to keep nfs up during storage cutover  using 7mtt? like CIFS

[SpindleNinja]

you would want zero data change during the final cutover. 

[san456]

I agree. However, like CIFS, NFS access cannot be terminated from storage side. When you have hundreds of NFS clients, and storage team do not have access to unmount them. We can only ask the users to unmount and not write any data during maintenance window.  Some may fail to do it.  I assume when 7mtt performs final sync, even if some writes are occurring by NFS clients to volume being transitioned, there will be a pause to complete the final sync and 7mtt will take the volume forcefully offline as part of the cutover. Therefore, users won’t be able to access 7 mode volumes.  If NFS clients use HARD mount option, it continues to retry to connect to NFS server when it becomes unavailable and successfully connects it when NFS server becomes available with same IP on cluster SVM after cutover complete.  Please share your thoughts

Have you run into a situation where NFS client freeze or encounter stale NFS handle that would require reboot of NFS clients to get back access to NFS shares after storage cutover to cluster SVM from 7 mode using 7MTT ?

Is there a better way to prevent NFS writes/changes during storage cutover form the storage side? Would you recommend making NFS permission for the volume being transitioned read-only before executing cutover using 7MTT? Not sure will this affect the cutover using 7MTT in any way.

[SpindleNinja]

I've checked through a lot of netapp docs as well as some past 7mtts i've done.   There's nothing that I can find that calls out hard vs soft specifically.     I can't recall either a time where a sever needed to be rebooted after/during cut over.   

Are you able to test it out in your environment? 

[san456]

Unfortunately NO. We do not have LAB or dev environment to test it out before production cutover.will try setup simulator and test it. 

[SpindleNinja]

Good luck on the 7MTT!

[moep]

You don't need to have NFS or CIFS running during the final "Storage Cutover" phase. It is sufficient to have it running in the Precutover phase for the "Apply Configuration" step. Once the CIFS and NFS configuration is replicated to the CDOT system you can stop NFS and CIFS on the source.

[san456]

Got it. Thank you. I was under impression that CIFS and NFS must be up and running during storage cutover. now it is clear.