There are scripts that have been written for this, but it's also something you shouldn't need to do.
If the host crashes, it ought to clear any of its old locks when remounting the filesystem. I'm not sure why some NFS shops run into this problem. I came to NetApp from Oracle where a developer would manage to crash a system at least daily, and in 7 years I never once had to clear locks. Not one time. I think the reason is related to name resolution. For example, if the host tried to clear locks as "hostname" but the locks were registered as "hostname.domain.com" then it would fail to clear the locks properly. My environments were rigidly based on short name all drawn from the same NIS-based hosts file, and we always had the same nsswitch.conf file that was ordered files, NIS, DNS.
I hear of a customer running into locking issues every 3 or 4 months and I've mentioned having a look at name resolution, but I've never had any conclusive feedback on whether they fixed the problem.
From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of Craig A. Falls Sent: Friday, April 06, 2012 7:30 PM To: toasters@teaparty.net Subject: NFS Lock and Oracle
Hi,
I was wondering if anyone had a script for viewing/removing NFS locks on a filer?
We have just started to use Oracle on NFS and have had an issue where we have had the host crash and leave locks which then stop the DB being started again (on that node) when that node is recovered and will probably (guessing) eventually stall the DB when other nodes try to access the data that is being locked.
I was planning on something like:
- check the locks for a filer/vfiler for the specific host - remove the locks
using lock status/lock break
Any thoughts/help or advice would be appreciated.
thanks c