Stale NFS mount point hanging bdf command

[John Goetz]

Is there an option when setting up a NFS mount point that if it no longer responds it will not hand a bdf command or various other commands?

[Paul Eadington_1 · Accepted solution]

The option your looking for is -o soft, this will return an error.

The default is hard which continues to rety until it gets a response.

[A. Clay Stephenson]

You do need to understand the implications of hard vs. soft NFS mounting. Fundamentally, under hard mounting NFS operations are just like their local counterparts and that is why hard is the default. Under soft mounting, the i/o operations can fail and report the status to the client. The application software may or may not be able to handle these errors gracefully. Under hard mounts, the i/o requests block until satisfied but under soft mounts the i/o operations fail and the application must know how to then handle the failure.

[Dave Olker]

Hi John,

As for your specific problem of bdf stops responding, have you checked the recent ONC patches for possible fixes to this problem? A quick check of the patch catalog showed this hit in PHNE_31097:


( SR:8606364383 CR:JAGaf25022 )
If an NFS server is not available, processes accessing NFS filesystems stops responding and cannot be killed.


The problem reported by the customer for this defect was "I do a bdf when the server is down and the bdf command stops responding. This causes other applications to stops responding." Sound familiar? The solution was to change how the NFS client handles locking data structures and sleeping at non-interruptible levels in the kernel.

However, even if you install the very latest ONC patch for 11.11 (PHNE_33315) you would see the same behavior. That's because the fix for this defect is disabled by default. The fix involved us creating a new adb kernel tunable parameter that changes this behavior and makes these non-interruptible (i.e. hung) processes interruptible (i.e. killable) again. The parameter name is "nfs_new_rnode_lock_code".

I'm not trying to suggest that this is the underlying cause of your problem, but you may want to investigate JAGaf25022 to see if it matches your situation.

Regards,

Dave

[Christine Hartman]

we run into this at times when the NFS server is down....it usually will restore the connection when that server comes back....in the meantime...if you use bdf -l instead of just bdf....it will only report locally mounted filesystems (not nfs mounts) and won't hang.
another option you might investigate is automounting...but that too does have it's shortcomings as well.

[Ramesh_Bhagat]

in case bdf got stuck ,  mijority of cases its because of NFS shares which are having issues . 

you can use bdf -l to see all local file systems & if bdf executed wuthout issue but bsf still hanging , its NFS shares issue for sure

to see & clear nfs shares ,

  run  mount -p    (to share all file systems)

umount -f  <nfs share>   (to forvefully umount nfs file system & clear bdf hanging issue)