NFSv4 stuck

[Rick Macklem]

Slawa Olhovchenkov wrote:

>On Thu, Jan 12, 2017 at 01:13:58PM +0700, Eugene Grosbein wrote:
>
>> 12.01.2017 6:50, Slawa Olhovchenkov пишет:
>> > On Wed, Jan 11, 2017 at 03:20:22PM -0800, Michael Sinatra wrote:
>> >
>> >> On 01/11/17 14:59, Slawa Olhovchenkov wrote:
>> >>> On Wed, Jan 11, 2017 at 10:39:42PM +0000, Rick Macklem wrote:
>> >>>
>> >>>> "umount -f" is your only chance. However, if there is already a non-forced
>> >>>> dismount stuck, it won't work because the non-forced dismount will have
>> >>>> the mounted-on vnode locked.
>> >>>
>> >>> I am use 'intr,soft' flags -- why ignored?
>> >>
>> >> intr and soft are considered dangerous with nfsv4.  See the "BUGS"
>> >> section of the mount_nfs(8) man page.
>> >
>> > Currently I am see dangerous w/o intr: system need reboot, application
>> > don't killable.
>>
>> NFSv4 confiders "intr,soft" dangerous for data generated by poorly written applications, not for a OS.
>
>What data? In may case no data.
Certain NFSv4 operations (such as open and byte range locking) are strictly ordered using a
seqid#. If you fail an RPC in progress (via a soft timeout or intr via a signal) then this seqid gets
out of sync between client and server and your mount is badly broken.

I do not believe this caused your hang though, since processes were sleeping on rpccon, which
means they were trying to do a new TCP connection to the server unsuccessfully.
- Which normally indicates a problem with your underlying network fabric.
stuff snipped...

rick