Hello,
I am getting an error when I run "ceph osd set require_jewel_osds
--yes-i-really-mean-it".
Error ENOENT: unknown feature '--yes-i-really-mean-it'
So I ran, "ceph osd set require_jewel_osds", and got this error:
Error EPERM: not all up OSDs have CEPH_FEATURE_SERVER_JEWEL feature
I verified all OSDs were stopped with "/etc/init.d/ceph-osd.N stop".
Then verified each was down with "ceph osd down N". When setting them
down, each replied "osd.N is already down". I started one of the OSDs
on a host that was downgraded to 10.2.3-r2 I then attempted to set
"ceph osd set require_jewel_osds", and get the same error.
The log for the OSD is showing this error:
2017-11-28 17:40:08.928446 7f47b082f940 1
filestore(/var/lib/ceph/osd/ceph-1) upgrade
2017-11-28 17:40:08.928475 7f47b082f940 2 osd.1 0 boot
2017-11-28 17:40:08.928788 7f47b082f940 -1 osd.1 0 The disk uses
features unsupported by the executable.
2017-11-28 17:40:08.928810 7f47b082f940 -1 osd.1 0 ondisk features
compat={},rocompat={},incompat={1=initial feature set(~v.18),2=pginfo
object,3=object
locator,4=last_epoch_clean,5=categories,6=hobjectpool,7=biginfo,8=leveldbinfo,9=leveldblog,10=snapmapper,12=transaction
hints,13=pg meta object,14=explicit missing set,15=fastinfo pg
attr,16=deletes in missing set}
2017-11-28 17:40:08.928818 7f47b082f940 -1 osd.1 0 daemon features
compat={},rocompat={},incompat={1=initial feature set(~v.18),2=pginfo
object,3=object
locator,4=last_epoch_clean,5=categories,6=hobjectpool,7=biginfo,8=leveldbinfo,9=leveldblog,10=snapmapper,11=sharded
objects,12=transaction hints,13=pg meta object}
2017-11-28 17:40:08.928827 7f47b082f940 -1 osd.1 0 Cannot write to
disk! Missing features: compat={},rocompat={},incompat={14=explicit
missing set,15=fastinfo pg attr,16=deletes in missing set}
2017-11-28 17:40:08.929353 7f47b082f940 1 journal close
/dev/disk/by-partlabel/ceph-1
2017-11-28 17:40:08.930488 7f47b082f940 -1 ** ERROR: osd init failed:
(95) Operation not supported
So the OSD is not starting because of missing features. It does not
show up in "ceph features" output.
Ceph features output:
ceph features
2017-11-28 17:51:31.213636 7f6a2140a700 -1 WARNING: the following
dangerous and experimental features are enabled: btrfs
2017-11-28 17:51:31.223068 7f6a2140a700 -1 WARNING: the following
dangerous and experimental features are enabled: btrfs
"mon": {
"group": {
"features": "0x1ffddff8eea4fffb",
"release": "luminous",
"num": 4
}
},
"mds": {
"group": {
"features": "0x7fddff8ee84bffb",
"release": "jewel",
"num": 1
}
},
"client": {
"group": {
"features": "0x1ffddff8eea4fffb",
"release": "luminous",
"num": 4
I attempted to set require_jewel_osds with the MGRs stopped, and had
the same results.
Output from ceph tell osd.1 version. I get the same error from all OSDs.
# ceph tell osd.1 versions
Error ENXIO: problem getting command descriptions from osd.1
Any thoughts?
Cary
-Dynamic
On Tue, Nov 28, 2017 at 1:09 PM, Sage Weil <sage@xxxxxxxxxxxx> wrote:
> On Tue, 28 Nov 2017, Cary wrote:
>> I get this error when I try to start the OSD that has been downgraded
>> to 10.2.3-r2.
>>
>> 2017-11-28 03:42:35.989754 7fa5e6429940 1
>> filestore(/var/lib/ceph/osd/ceph-3) upgrade
>> 2017-11-28 03:42:35.989788 7fa5e6429940 2 osd.3 0 boot
>> 2017-11-28 03:42:35.990132 7fa5e6429940 -1 osd.3 0 The disk uses
>> features unsupported by the executable.
>> 2017-11-28 03:42:35.990142 7fa5e6429940 -1 osd.3 0 ondisk features
>> compat={},rocompat={},incompat={1=initial feature set(~v.18),2=pginfo
>> object,3=object
>> locator,4=last_epoch_clean,5=categories,6=hobjectpool,7=biginfo,8=leveldbinfo,9=leveldblog,10=snapmapper,12=transaction
>> hints,13=pg meta object,14=explicit missing set,15=fastinfo pg
>> attr,16=deletes in missing set}
>> 2017-11-28 03:42:35.990150 7fa5e6429940 -1 osd.3 0 daemon features
>> compat={},rocompat={},incompat={1=initial feature set(~v.18),2=pginfo
>> object,3=object
>> locator,4=last_epoch_clean,5=categories,6=hobjectpool,7=biginfo,8=leveldbinfo,9=leveldblog,10=snapmapper,11=sharded
>> objects,12=transaction hints,13=pg meta object}
>> 2017-11-28 03:42:35.990160 7fa5e6429940 -1 osd.3 0 Cannot write to
>> disk! Missing features: compat={},rocompat={},incompat={14=explicit
>> missing set,15=fastinfo pg attr,16=deletes in missing set}
>> 2017-11-28 03:42:35.990775 7fa5e6429940 1 journal close
>> /dev/disk/by-partlabel/ceph-3
>> 2017-11-28 03:42:35.992960 7fa5e6429940 -1 ** ERROR: osd init failed:
>> (95) Operation not supported
>
> Oh, right. In that case, install the 'luminous' branch[1] on the monitors
> (or just the primary monitor if you're being conservative), restrart it,
> and you'll be able to do
>
> ceph osd set require_jewel_osds --yes-i-really-mean-it
>
> sage
>
>
> [1] ceph-deploy install --dev luminous HOST
>
>
>
>
>> Cary
>>
>> On Tue, Nov 28, 2017 at 3:09 AM, Sage Weil <sage@xxxxxxxxxxxx> wrote:
>> > On Tue, 28 Nov 2017, Cary wrote:
>> >> Hello,
>> >>
>> >> Could someone please help me complete my botched upgrade from Jewel
>> >> 10.2.3-r1 to Luminous 12.2.1. I have 9 Gentoo servers, 4 of which have
>> >> 2 OSDs each.
>> >>
>> >> My OSD servers were accidentally rebooted before the monitor servers
>> >> causing them to be running Luminous before the monitors. All services
>> >> have been restarted and running ceph versions gives the following:
>> >>
>> >> # ceph versions
>> >> 2017-11-27 21:27:24.356940 7fed67efe700 -1 WARNING: the following
>> >> dangerous and experimental features are enabled: btrfs
>> >> 2017-11-27 21:27:24.368469 7fed67efe700 -1 WARNING: the following
>> >> dangerous and experimental features are enabled: btrfs
>> >>
>> >> "mon": {
>> >> "ceph version 12.2.1
>> >> (3e7492b9ada8bdc9a5cd0feafd42fbca27f9c38e) luminous (stable)": 4
>> >> },
>> >> "mgr": {
>> >> "ceph version 12.2.1
>> >> (3e7492b9ada8bdc9a5cd0feafd42fbca27f9c38e) luminous (stable)": 3
>> >> },
>> >> "osd": {},
>> >> "mds": {
>> >> "ceph version 12.2.1
>> >> (3e7492b9ada8bdc9a5cd0feafd42fbca27f9c38e) luminous (stable)": 1
>> >> },
>> >> "overall": {
>> >> "ceph version 12.2.1
>> >> (3e7492b9ada8bdc9a5cd0feafd42fbca27f9c38e) luminous (stable)": 8
>> >>
>> >>
>> >>
>> >> For some reason the OSDs do not show what version they are running,
>> >> and a ceph osd tree shows all of the OSD as being down.
>> >>
>> >> # ceph osd tree
>> >> 2017-11-27 21:32:51.969335 7f483d9c2700 -1 WARNING: the following
>> >> dangerous and experimental features are enabled: btrfs
>> >> 2017-11-27 21:32:51.980976 7f483d9c2700 -1 WARNING: the following
>> >> dangerous and experimental features are enabled: btrfs
>> >> ID CLASS WEIGHT TYPE NAME STATUS REWEIGHT PRI-AFF
>> >> -1 27.77998 root default
>> >> -3 27.77998 datacenter DC1
>> >> -6 27.77998 rack 1B06
>> >> -5 6.48000 host ceph3
>> >> 1 1.84000 osd.1 down 0 1.00000
>> >> 3 4.64000 osd.3 down 0 1.00000
>> >> -2 5.53999 host ceph4
>> >> 5 4.64000 osd.5 down 0 1.00000
>> >> 8 0.89999 osd.8 down 0 1.00000
>> >> -4 9.28000 host ceph6
>> >> 0 4.64000 osd.0 down 0 1.00000
>> >> 2 4.64000 osd.2 down 0 1.00000
>> >> -7 6.48000 host ceph7
>> >> 6 4.64000 osd.6 down 0 1.00000
>> >> 7 1.84000 osd.7 down 0 1.00000
>> >>
>> >> The OSD logs all have this message:
>> >>
>> >> 20235 osdmap REQUIRE_JEWEL OSDMap flag is NOT set; please set it.
>> >
>> > THis is an annoying corner condition. 12.2.2 (out soon!) will have a
>> > --force option to set the flag even tho no osds are up. Until then, the
>> > workaround is to downgrade one host to jewel, start one jewel osd, then
>> > set the flag. Then upgrade to luminous again and restart all osds.
>> >
>> > sage
>> >
>> >
>> >>
>> >> When I try to set it with "ceph osd set require_jewel_osds" I get this error:
>> >>
>> >> Error EPERM: not all up OSDs have CEPH_FEATURE_SERVER_JEWEL feature
>> >>
>> >>
>> >>
>> >> A "ceph features" returns:
>> >>
>> >> "mon": {
>> >> "group": {
>> >> "features": "0x1ffddff8eea4fffb",
>> >> "release": "luminous",
>> >> "num": 4
>> >> }
>> >> },
>> >> "mds": {
>> >> "group": {
>> >> "features": "0x1ffddff8eea4fffb",
>> >> "release": "luminous",
>> >> "num": 1
>> >> }
>> >> },
>> >> "osd": {
>> >> "group": {
>> >> "features": "0x1ffddff8eea4fffb",
>> >> "release": "luminous",
>> >> "num": 8
>> >> }
>> >> },
>> >> "client": {
>> >> "group": {
>> >> "features": "0x1ffddff8eea4fffb",
>> >> "release": "luminous",
>> >> "num": 3
>> >>
>> >> # ceph tell osd.* versions
>> >> 2017-11-28 02:29:28.565943 7f99c6aee700 -1 WARNING: the following
>> >> dangerous and experimental features are enabled: btrfs
>> >> 2017-11-28 02:29:28.578956 7f99c6aee700 -1 WARNING: the following
>> >> dangerous and experimental features are enabled: btrfs
>> >> Error ENXIO: problem getting command descriptions from osd.0
>> >> osd.0: problem getting command descriptions from osd.0
>> >> Error ENXIO: problem getting command descriptions from osd.1
>> >> osd.1: problem getting command descriptions from osd.1
>> >> Error ENXIO: problem getting command descriptions from osd.2
>> >> osd.2: problem getting command descriptions from osd.2
>> >> Error ENXIO: problem getting command descriptions from osd.3
>> >> osd.3: problem getting command descriptions from osd.3
>> >> Error ENXIO: problem getting command descriptions from osd.5
>> >> osd.5: problem getting command descriptions from osd.5
>> >> Error ENXIO: problem getting command descriptions from osd.6
>> >> osd.6: problem getting command descriptions from osd.6
>> >> Error ENXIO: problem getting command descriptions from osd.7
>> >> osd.7: problem getting command descriptions from osd.7
>> >> Error ENXIO: problem getting command descriptions from osd.8
>> >> osd.8: problem getting command descriptions from osd.8
>> >>
>> >> # ceph daemon osd.1 status
>> >>
>> >> "cluster_fsid": "CENSORED",
>> >> "osd_fsid": "CENSORED",
>> >> "whoami": 1,
>> >> "state": "preboot",
>> >> "oldest_map": 19482,
>> >> "newest_map": 20235,
>> >> "num_pgs": 141
>> >>
>> >> # ceph -s
>> >> 2017-11-27 22:04:10.372471 7f89a3935700 -1 WARNING: the following
>> >> dangerous and experimental features are enabled: btrfs
>> >> 2017-11-27 22:04:10.375709 7f89a3935700 -1 WARNING: the following
>> >> dangerous and experimental features are enabled: btrfs
>> >> cluster:
>> >> id: CENSORED
>> >> health: HEALTH_ERR
>> >> 513 pgs are stuck inactive for more than 60 seconds
>> >> 126 pgs backfill_wait
>> >> 52 pgs backfilling
>> >> 435 pgs degraded
>> >> 513 pgs stale
>> >> 435 pgs stuck degraded
>> >> 513 pgs stuck stale
>> >> 435 pgs stuck unclean
>> >> 435 pgs stuck undersized
>> >> 435 pgs undersized
>> >> recovery 854719/3688140 objects degraded (23.175%)
>> >> recovery 838607/3688140 objects misplaced (22.738%)
>> >> mds cluster is degraded
>> >> crush map has straw_calc_version=0
>> >>
>> >> services:
>> >> mon: 4 daemons, quorum 0,1,3,2
>> >> mgr: 0(active), standbys: 1, 5
>> >> mds: cephfs-1/1/1 up {0=a=up:replay}, 1 up:standby
>> >> osd: 8 osds: 0 up, 0 in
>> >>
>> >> data:
>> >> pools: 7 pools, 513 pgs
>> >> objects: 1199k objects, 4510 GB
>> >> usage: 13669 GB used, 15150 GB / 28876 GB avail
>> >> pgs: 854719/3688140 objects degraded (23.175%)
>> >> 838607/3688140 objects misplaced (22.738%)
>> >> 257 stale+active+undersized+degraded
>> >> 126 stale+active+undersized+degraded+remapped+backfill_wait
>> >> 78 stale+active+clean
>> >> 52 stale+active+undersized+degraded+remapped+backfilling
>> >>
>> >>
>> >> I ran "ceph auth list", and client.admin has the following permissions.
>> >> auid: 0
>> >> caps: [mds] allow
>> >> caps: [mgr] allow *
>> >> caps: [mon] allow *
>> >> caps: [osd] allow *
>> >>
>> >> Thank you for your time.
>> >>
>> >> Is there any way I can get these OSDs to join the cluster now, or
>> >> recover my data?
>> >>
>> >> Cary
>> >> -Dynamic
>> >> --
>> >> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
>> >> the body of a message to majordomo@xxxxxxxxxxxxxxx
>> >> More majordomo info at http://vger.kernel.org/majordomo-info.html
>> >>
>> >>
>>
>>
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
