Re: pNFS timeouts

[Benny Halevy <bhalevy@panasas.com>]

On Jun. 14, 2010, 11:05 -0400, Lukas Hejtmanek <xhejtman@ics.muni.cz> wrote:
> On Mon, Jun 14, 2010 at 10:46:55AM -0400, Benny Halevy wrote:
>>> Jun 14 16:17:00 undomiel1 kernel: [187434.936507] nfs41_sequence_call_done rpc_cred ffff88003d7a0300
>>> Jun 14 16:17:00 undomiel1 kernel: [187434.936514] <-- nfs41_sequence_call_done
>>> Jun 14 16:17:00 undomiel1 kernel: [187434.936522] nfs4_schedule_state_renewal: requeueing work. Lease period = 60
>>> Jun 14 16:17:00 undomiel1 kernel: [187434.936530] --> nfs_put_client({2})
>>> Jun 14 16:17:16 undomiel1 kernel: [187450.536014] nfs4_renew_state: start
>>> Jun 14 16:17:16 undomiel1 kernel: [187450.536045] nfs4_renew_state: failed to call renewd. Reason: lease not expired
>>> Jun 14 16:17:16 undomiel1 kernel: [187450.536056] nfs4_schedule_state_renewal: requeueing work. Lease period = 38
>>
>> This looks like a complaint but it's actually saying that the "failure"
>> is due to the fact that everything is OK (lease not expired)
> 
> the state renewal actually starts after 16 seconds after schedule start
> renewal for some reason. The nfs_put_client actually starts the renewal?
> 

No, the state renewal is done independently in the background.

>> BTW, is this the same issue that Jiri reported?
>> http://marc.info/?l=linux-nfs&m=127619696004097&w=2
> 
> Yes, but I didn't see any response to that, so I guess his report is not clear
> enough.
>