985461 - Use 'nolock' NFS option for readonly mounts

Status: RESOLVED FIXED

(Infrastructure & Operations :: Storage, task)

Description

[Greg Cox [:gcox]]

Due to a misbehavior with NetApp ONTAP (seen in 8.2 cDOT), an NFS client who establishes a read lock against a readonly volume creates a lock that doesn't clear when called to unlock.  This 'lock leak' pileup can get in the way of nondisruptive operations like volume moves.

Adding 'nolock' as an NFS option should alleviate the issue, while not causing any side issues (since having the (default) locks are advisory at best, and stupid at worst).

I've advised NetApp of this.  If there's a BURT opened against it later, I'll note it in this bug.

TODO:
* Doc this in our volume creations/export change guidelines.
* Audit where we're running afoul of this, open sub-bugs to fix them.
* Clean up residual locks.

Comment 1

[Greg Cox [:gcox]]

Closed out our netapp case, BURT 756081.  The symptoms don't quite sound exactly right, but since we're past the issue via nolock, it's not so much of a worry.

That BURT is resolved in the recently-released 8.2P6 and 8.2.1.