Thanks Sage! What had happened prior to me upgrading was that I added an
erasure coded pool, but all my OSDs began to crash. The ec profile didn't
seem to cause the crash, so I left it, but once I removed the pool, the
crashes stopped.
Do you guys want any of the core dumps, or is anything short of a debug
build with symbols and gdb on my system going to help?
Joshua
On Fri, Jul 11, 2014 at 7:02 AM, Sage Weil <sweil at redhat.com> wrote:
> On Thu, 10 Jul 2014, Joshua McClintock wrote:
> > { "rule_id": 1,
> > "rule_name": "erasure-code",
> > "ruleset": 1,
> > "type": 3,
>
> The presence of the erasure code CRUSH rules it what is preventing the
> kernel client from mounting. Upgrade to a newer kernel (3.14 I think?),
> or remove the EC pools and these CRUSH rules to allow the old client to
> mount.
>
> > "tunables": { "choose_local_tries": 0,
> > "choose_local_fallback_tries": 0,
> > "choose_total_tries": 50,
> > "chooseleaf_descend_once": 1,
> > "profile": "bobtail",
> > "optimal_tunables": 0,
> > "legacy_tunables": 0,
> > "require_feature_tunables": 1,
> > "require_feature_tunables2": 1}}
>
> Looks like fields for tunables3 and the v2 and v3 rules are missing from
> the dump; I'll fix that.
>
> Thanks!
> sage
>
>
> >
> >
> >
> > On Thu, Jul 10, 2014 at 8:16 PM, Sage Weil <sweil at redhat.com> wrote:
> > That is CEPH_FEATURE_CRUSH_V2. Can you attach teh output of
> >
> > ceph osd crush dump
> >
> > Thanks!
> > sage
> >
> >
> > On Thu, 10 Jul 2014, Joshua McClintock wrote:
> >
> > > Yes, I change some of the mount options on my osds (xfs mount
> > options), but
> > > I think this may be the answer from dmesg, sorta looks like a
> > version
> > > mismatch:
> > >
> > > libceph: loaded (mon/osd proto 15/24)
> > >
> > > ceph: loaded (mds proto 32)
> > >
> > > libceph: mon0 192.168.0.14:6789 feature set mismatch, my
> > 4a042aca < server's
> > > 104a042aca, missing 1000000000
> > >
> > > libceph: mon0 192.168.0.14:6789 socket error on read
> > >
> > > libceph: mon2 192.168.0.16:6789 feature set mismatch, my
> > 4a042aca < server's
> > > 104a042aca, missing 1000000000
> > >
> > > libceph: mon2 192.168.0.16:6789 socket error on read
> > >
> > > libceph: mon1 192.168.0.15:6789 feature set mismatch, my
> > 4a042aca < server's
> > > 104a042aca, missing 1000000000
> > >
> > > libceph: mon1 192.168.0.15:6789 socket error on read
> > >
> > > libceph: mon0 192.168.0.14:6789 feature set mismatch, my
> > 4a042aca < server's
> > > 104a042aca, missing 1000000000
> > >
> > > libceph: mon0 192.168.0.14:6789 socket error on read
> > >
> > > libceph: mon2 192.168.0.16:6789 feature set mismatch, my
> > 4a042aca < server's
> > > 104a042aca, missing 1000000000
> > >
> > > libceph: mon2 192.168.0.16:6789 socket error on read
> > >
> > > libceph: mon1 192.168.0.15:6789 feature set mismatch, my
> > 4a042aca < server's
> > > 104a042aca, missing 1000000000
> > >
> > > libceph: mon1 192.168.0.15:6789 socket error on read
> > >
> > >
> > > I maybe I didn't update as well as I thought it did. I did
> > hit every mon,
> > > but I remember I couldn't upgrade to the new 'ceph' package
> > because it
> > > conflicted with 'python-ceph', so I uninstalled it
> > (python-ceph), and then
> > > upgraded to .80.1-2. Maybe there's a subcomponent I missed?
> > >
> > >
> > > Here's rpm -qa from the client:
> > >
> > >
> > > [root at chefwks01 ~]# rpm -qa|grep ceph
> > >
> > > ceph-deploy-1.5.2-0.noarch
> > >
> > > ceph-release-1-0.el6.noarch
> > >
> > > ceph-0.80.1-2.el6.x86_64
> > >
> > > libcephfs1-0.80.1-0.el6.x86_64
> > >
> > >
> > > Here's rpm -qa from the mons:
> > >
> > >
> > > [root at ceph-mon01 ~]# rpm -qa|grep ceph
> > >
> > > ceph-0.80.1-2.el6.x86_64
> > >
> > > ceph-release-1-0.el6.noarch
> > >
> > > libcephfs1-0.80.1-0.el6.x86_64
> > >
> > > [root at ceph-mon01 ~]#
> > >
> > >
> > > [root at ceph-mon02 ~]# rpm -qa|grep ceph
> > >
> > > libcephfs1-0.80.1-0.el6.x86_64
> > >
> > > ceph-0.80.1-2.el6.x86_64
> > >
> > > ceph-release-1-0.el6.noarch
> > >
> > > [root at ceph-mon02 ~]#
> > >
> > >
> > > [root at ceph-mon03 ~]# rpm -qa|grep ceph
> > >
> > > libcephfs1-0.80.1-0.el6.x86_64
> > >
> > > ceph-0.80.1-2.el6.x86_64
> > >
> > > ceph-release-1-0.el6.noarch
> > >
> > > [root at ceph-mon03 ~]#
> > >
> > >
> > > Joshua
> > >
> > >
> > >
> > > On Thu, Jul 10, 2014 at 6:04 PM, Sage Weil <sweil at redhat.com>
> > wrote:
> > > Have you made any other changes after the upgrade?
> > (Like
> > > adjusting
> > > tunables, or creating EC pools?)
> > >
> > > See if there is anything in 'dmesg' output.
> > >
> > > sage
> > >
> > > On Thu, 10 Jul 2014, Joshua McClintock wrote:
> > >
> > > > I upgraded my cluster to .80.1-2 (CentOS). My mount
> > command
> > > just freezes
> > > > and outputs an error:
> > > >
> > > > mount.ceph 192.168.0.14,192.168.0.15,192.168.0.16:/
> > /us-west01
> > > -o
> > > > name=chefwks01,secret=`ceph-authtool -p -n
> > client.admin
> > > > /etc/ceph/us-west01.client.admin.keyring`
> > > >
> > > > mount error 5 = Input/output error
> > > >
> > > >
> > > > Here's the output from 'ceph -s'
> > > >
> > > >
> > > > cluster xxxxxxxxxxxxxxxxxxxxxx
> > > >
> > > > health HEALTH_OK
> > > >
> > > > monmap e1: 3monsat{ceph-mon01=
> 192.168.0.14:6789/0,ceph-mon02=192.168.0.15:6789/0,ceph-m
> > on03
> > > =1
> > > > 92.168.0.16:6789/0}, election epoch 88, quorum 0,1,2
> > > > ceph-mon01,ceph-mon02,ceph-mon03
> > > >
> > > > mdsmap e26: 1/1/1 up {0=0=up:active}
> > > >
> > > > osdmap e1371: 5 osds: 5 up, 5 in
> > > >
> > > > pgmap v49431: 192 pgs, 3 pools, 135 GB data, 34733
> > > objects
> > > >
> > > > 406 GB used, 1874 GB / 2281 GB avail
> > > >
> > > > 192 active+clean
> > > >
> > > >
> > > > I can see some packets being exchanged between the client
> > and
> > > the mon, but
> > > > it's a pretty short exchange.
> > > >
> > > > Any ideas where to look next?
> > > >
> > > > Joshua
> > > >
> > > >
> > > >
> > >
> > >
> > >
> > >
> >
> >
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ceph.com/pipermail/ceph-users-ceph.com/attachments/20140711/5126f747/attachment.htm>
