Guys, thanks for looking into this, turns out I had the following options configured on the MDS:
mds_session_timeout = 10
mds_session_autoclose = 15
I must have set them when fiddling around with some ctdb stuff. After reverting to the defaults (60 and 300 respectively), I'm no longer seeing the stale caps errors.
Sorry for the noise.
Thanks,
On Sat, Sep 3, 2016 at 4:14 AM, Yan, Zheng <ukernel@xxxxxxxxx> wrote:
On Sat, Sep 3, 2016 at 1:35 AM, Gregory Farnum <gfarnum@xxxxxxxxxx> wrote:
> On Fri, Sep 2, 2016 at 2:58 AM, David <dclistslinux@xxxxxxxxx> wrote:
>> Hi All
>>
>> Kernel client: 4.6.4-1.el7.elrepo.x86_64
>> MDS version: 10.2.2
>> OS: CentOS 7
>>
>> I have Cephfs mounted on a few servers, I see the following in the log
>> approx every 20 secs on all of them:
>>
>> kernel: ceph: mds0 caps went stale, renewing
>> kernel: ceph: mds0 caps stale
>> kernel: ceph: mds0 caps renewed
>>
>> I'm trying to debug a few intermittent nfs related issues (which may or may
>> not be cephfs related) and I'm just wondering if these lines are anything to
>> worry about.
>>
>> I've mounted on a server that is almost completely idle and I'm still seeing
>> those lines every 20 secs.
>
> If it's idle I think that's why — normally these are kept live
> (unstale) just by passing normal messages back and forth. Although,
> Zheng, it ought to be sending off messages prior to stale if there's
> no other traffic, shouldn't it?
client sends CEPH_SESSION_REQUEST_RENEWCAPS to mds even it's idle.
Regards
Yan, Zheng
>
> If you're not seeing any problems on these hosts I don't think you
> need to worry about it, though.
>
>
>> Ceph health is OK and the MDS server seems pretty happy although I am
>> occasionally seeing some "closing stale session" lines in the ceph-mds log
>> but I think that's a separate issue.
>
> Closing stale sessions means the MDS is removing clients which haven't
> checked in for a while. That might be a problem if the client computer
> is still around and wants to access the FS. (Alternatively, if you've
> got eg laptops mounting CephFS and they're timing out when they get
> taken home for the night, not so much.)
> -Greg
> _______________________________________________
> ceph-users mailing list
> ceph-users@xxxxxxxxxxxxxx
> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph. com
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph. com
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
