ceph Nautilus lost two disk over night everything hangs
Hello, i run a ceph Nautilus cluster with 9 hosts and 144 OSDs. Last night we lost two disks, so two OSDs (67,90) are down. The two disks are on two different hosts. A third ODS on a third host repotrts slow ops. ceph is repairing at the moment. Pools affected are eg these ones: pool 35 'pxa-rbd' replicated size 3 min_size 2 crush_rule 0 object_hash rjenkins pg_num 256 pgp_num 256 last_change 192082 lfor 0/27841/27845 flags hashpspool,selfmanaged_snaps stripe_width 0 pg_num_min 128 target_size_ratio 0.0001 application rbd pool 36 'pxa-ec' erasure size 6 min_size 5 crush_rule 7 object_hash rjenkins pg_num 512 pgp_num 512 last_change 192177 lfor 0/172580/172578 flags hashpspool,ec_overwrites,selfmanaged_snaps stripe_width 16384 pg_num_min 512 target_size_ratio 0.15 application rbd At the mmoment the proxmox-cluster using storage from the seperate ceph cluster hangs. The ppols with date are erasure coded with the following profile: crush-device-class= crush-failure-domain=host crush-root=default jerasure-per-chunk-alignment=false k=4 m=2 plugin=jerasure technique=reed_sol_van w=8 What I do not understand is why access on the virtualization seem to block. Could that be related to min_size of the pools cause this behaviour? How can I find out if this is true or what else is causing the blocking behaviour seen? This is the current status: health: HEALTH_WARN Reduced data availability: 1 pg inactive, 1 pg incomplete Degraded data redundancy: 42384/130014984 objects degraded (0.033%), 4 pgs degraded, 5 pgs undersized 15 daemons have recently crashed 150 slow ops, oldest one blocked for 15901 sec, daemons [osd.60,osd.67] have slow ops. services: mon: 3 daemons, quorum ceph2,ceph5,ceph8 (age 4h) mgr: ceph2(active, since 7w), standbys: ceph5, ceph8, ceph-admin mds: cephfsrz:1 {0=ceph6=up:active} 2 up:standby osd: 144 osds: 142 up (since 4h), 142 in (since 5h); 6 remapped pgs task status: scrub status: mds.ceph6: idle data: pools: 15 pools, 2632 pgs objects: 21.70M objects, 80 TiB usage: 139 TiB used, 378 TiB / 517 TiB avail pgs: 0.038% pgs not active 42384/130014984 objects degraded (0.033%) 2623 active+clean 3 active+undersized+degraded+remapped+backfilling 3 active+clean+scrubbing+deep 1 active+undersized+degraded+remapped+backfill_wait 1 active+undersized+remapped+backfill_wait 1 remapped+incomplete io: client: 2.2 MiB/s rd, 3.6 MiB/s wr, 8 op/s rd, 179 op/s wr recovery: 51 MiB/s, 12 objects/s Thanks a lot Rainer -- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312
Hi, from what you've sent my conclusion about the stalled I/O would be indeed the min_size of the EC pool. There's only one PG reported as incomplete, I assume that is the EC pool, not the replicated pxa-rbd, right? Both pools are for rbd so I'm guessing the rbd headers are in pxa-rbd while the data is stored in pxa-ec, could you confirm that? You could add 'ceph health detail' output to your question to see which PG is incomplete. I assume that both down OSDs are in the acting set of the inactive PG, and since the pool's min_size is 5 the I/O pauses. If you can't wait for recovery to finish and can't bring up at least one of those OSDs you could set the min_size of pxa-ec to 4, but if you do, be aware that one more disk failure could mean data loss! So think carefully about it (maybe you could instead speed up recovery?) and don't forget to increase min_size back to 5 when the recovery has finished, that's very important! Regards, Eugen Zitat von Rainer Krienke <krienke@uni-koblenz.de>:
Hello,
i run a ceph Nautilus cluster with 9 hosts and 144 OSDs. Last night we lost two disks, so two OSDs (67,90) are down. The two disks are on two different hosts. A third ODS on a third host repotrts slow ops. ceph is repairing at the moment.
Pools affected are eg these ones: pool 35 'pxa-rbd' replicated size 3 min_size 2 crush_rule 0 object_hash rjenkins pg_num 256 pgp_num 256 last_change 192082 lfor 0/27841/27845 flags hashpspool,selfmanaged_snaps stripe_width 0 pg_num_min 128 target_size_ratio 0.0001 application rbd
pool 36 'pxa-ec' erasure size 6 min_size 5 crush_rule 7 object_hash rjenkins pg_num 512 pgp_num 512 last_change 192177 lfor 0/172580/172578 flags hashpspool,ec_overwrites,selfmanaged_snaps stripe_width 16384 pg_num_min 512 target_size_ratio 0.15 application rbd
At the mmoment the proxmox-cluster using storage from the seperate ceph cluster hangs. The ppols with date are erasure coded with the following profile:
crush-device-class= crush-failure-domain=host crush-root=default jerasure-per-chunk-alignment=false k=4 m=2 plugin=jerasure technique=reed_sol_van w=8
What I do not understand is why access on the virtualization seem to block. Could that be related to min_size of the pools cause this behaviour? How can I find out if this is true or what else is causing the blocking behaviour seen?
This is the current status: health: HEALTH_WARN Reduced data availability: 1 pg inactive, 1 pg incomplete Degraded data redundancy: 42384/130014984 objects degraded (0.033%), 4 pgs degraded, 5 pgs undersized 15 daemons have recently crashed 150 slow ops, oldest one blocked for 15901 sec, daemons [osd.60,osd.67] have slow ops.
services: mon: 3 daemons, quorum ceph2,ceph5,ceph8 (age 4h) mgr: ceph2(active, since 7w), standbys: ceph5, ceph8, ceph-admin mds: cephfsrz:1 {0=ceph6=up:active} 2 up:standby osd: 144 osds: 142 up (since 4h), 142 in (since 5h); 6 remapped pgs
task status: scrub status: mds.ceph6: idle
data: pools: 15 pools, 2632 pgs objects: 21.70M objects, 80 TiB usage: 139 TiB used, 378 TiB / 517 TiB avail pgs: 0.038% pgs not active 42384/130014984 objects degraded (0.033%) 2623 active+clean 3 active+undersized+degraded+remapped+backfilling 3 active+clean+scrubbing+deep 1 active+undersized+degraded+remapped+backfill_wait 1 active+undersized+remapped+backfill_wait 1 remapped+incomplete
io: client: 2.2 MiB/s rd, 3.6 MiB/s wr, 8 op/s rd, 179 op/s wr recovery: 51 MiB/s, 12 objects/s
Thanks a lot Rainer -- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312 _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
Hello, yes your assumptions are correct pxa-rbd ist the metadata pool for pxa-ec which uses a erasure coding 4+2 profile. In the last hours ceph repaired most of the damage. One inactive PG remained and in ceph health detail then told me: --------- HEALTH_WARN Reduced data availability: 1 pg inactive, 1 pg incomplete; 15 daemons have recently crashed; 150 slow ops, oldest one blocked for 26716 sec, daemons [osd.60,osd.67] have slow ops. PG_AVAILABILITY Reduced data availability: 1 pg inactive, 1 pg incomplete pg 36.15b is remapped+incomplete, acting [60,2147483647,23,96,2147483647,36] (reducing pool pxa-ec min_size from 5 may help; search ceph.com/docs for 'incomplete') RECENT_CRASH 15 daemons have recently crashed osd.90 crashed on host ceph6 at 2021-03-29 21:14:10.442314Z osd.67 crashed on host ceph5 at 2021-03-30 02:21:23.944205Z osd.67 crashed on host ceph5 at 2021-03-30 01:39:14.452610Z osd.90 crashed on host ceph6 at 2021-03-29 21:14:24.222223Z osd.67 crashed on host ceph5 at 2021-03-30 02:35:43.373845Z osd.67 crashed on host ceph5 at 2021-03-30 01:19:58.762393Z osd.67 crashed on host ceph5 at 2021-03-30 02:09:42.297941Z osd.67 crashed on host ceph5 at 2021-03-30 02:28:29.981528Z osd.67 crashed on host ceph5 at 2021-03-30 01:50:05.374278Z osd.90 crashed on host ceph6 at 2021-03-29 21:13:51.896849Z osd.67 crashed on host ceph5 at 2021-03-30 02:00:22.593745Z osd.67 crashed on host ceph5 at 2021-03-30 01:29:39.170134Z osd.90 crashed on host ceph6 at 2021-03-29 21:14:38.114768Z osd.67 crashed on host ceph5 at 2021-03-30 00:54:06.629808Z osd.67 crashed on host ceph5 at 2021-03-30 01:10:21.824447Z --------- All osds except for 67 and 90 are up and I followed the hint in health detail and lowered min_size from 5 to 4 for pxa-ec. Since then ceph is again repairing and in between some VMs in the attached proxmox cluster are working again. So I hope that after repairing all PGs are up, so that I can restart all VMs again. Thanks Rainer Am 30.03.21 um 11:41 schrieb Eugen Block:
Hi,
from what you've sent my conclusion about the stalled I/O would be indeed the min_size of the EC pool. There's only one PG reported as incomplete, I assume that is the EC pool, not the replicated pxa-rbd, right? Both pools are for rbd so I'm guessing the rbd headers are in pxa-rbd while the data is stored in pxa-ec, could you confirm that?
You could add 'ceph health detail' output to your question to see which PG is incomplete. I assume that both down OSDs are in the acting set of the inactive PG, and since the pool's min_size is 5 the I/O pauses. If you can't wait for recovery to finish and can't bring up at least one of those OSDs you could set the min_size of pxa-ec to 4, but if you do, be aware that one more disk failure could mean data loss! So think carefully about it (maybe you could instead speed up recovery?) and don't forget to increase min_size back to 5 when the recovery has finished, that's very important!
Regards, Eugen
Zitat von Rainer Krienke <krienke@uni-koblenz.de>:
Hello,
i run a ceph Nautilus cluster with 9 hosts and 144 OSDs. Last night we lost two disks, so two OSDs (67,90) are down. The two disks are on two different hosts. A third ODS on a third host repotrts slow ops. ceph is repairing at the moment.
Pools affected are eg these ones: pool 35 'pxa-rbd' replicated size 3 min_size 2 crush_rule 0 object_hash rjenkins pg_num 256 pgp_num 256 last_change 192082 lfor 0/27841/27845 flags hashpspool,selfmanaged_snaps stripe_width 0 pg_num_min 128 target_size_ratio 0.0001 application rbd
pool 36 'pxa-ec' erasure size 6 min_size 5 crush_rule 7 object_hash rjenkins pg_num 512 pgp_num 512 last_change 192177 lfor 0/172580/172578 flags hashpspool,ec_overwrites,selfmanaged_snaps stripe_width 16384 pg_num_min 512 target_size_ratio 0.15 application rbd
At the mmoment the proxmox-cluster using storage from the seperate ceph cluster hangs. The ppols with date are erasure coded with the following profile:
crush-device-class= crush-failure-domain=host crush-root=default jerasure-per-chunk-alignment=false k=4 m=2 plugin=jerasure technique=reed_sol_van w=8
What I do not understand is why access on the virtualization seem to block. Could that be related to min_size of the pools cause this behaviour? How can I find out if this is true or what else is causing the blocking behaviour seen?
This is the current status: health: HEALTH_WARN Reduced data availability: 1 pg inactive, 1 pg incomplete Degraded data redundancy: 42384/130014984 objects degraded (0.033%), 4 pgs degraded, 5 pgs undersized 15 daemons have recently crashed 150 slow ops, oldest one blocked for 15901 sec, daemons [osd.60,osd.67] have slow ops.
services: mon: 3 daemons, quorum ceph2,ceph5,ceph8 (age 4h) mgr: ceph2(active, since 7w), standbys: ceph5, ceph8, ceph-admin mds: cephfsrz:1 {0=ceph6=up:active} 2 up:standby osd: 144 osds: 142 up (since 4h), 142 in (since 5h); 6 remapped pgs
task status: scrub status: mds.ceph6: idle
data: pools: 15 pools, 2632 pgs objects: 21.70M objects, 80 TiB usage: 139 TiB used, 378 TiB / 517 TiB avail pgs: 0.038% pgs not active 42384/130014984 objects degraded (0.033%) 2623 active+clean 3 active+undersized+degraded+remapped+backfilling 3 active+clean+scrubbing+deep 1 active+undersized+degraded+remapped+backfill_wait 1 active+undersized+remapped+backfill_wait 1 remapped+incomplete
io: client: 2.2 MiB/s rd, 3.6 MiB/s wr, 8 op/s rd, 179 op/s wr recovery: 51 MiB/s, 12 objects/s
Thanks a lot Rainer -- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312 _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
-- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312
Hi, this is odd. The problem with recovery when sufficiently many but less than min_size shards are present should have been resolved with osd_allow_recovery_below_min_size=true. It is really dangerous to reduce min_size below k+1 and, in fact, should never be necessary for recovery. Can you check if this option is present and set to true? If it is not working as intended, a tracker ticker might be in order. Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14 ________________________________________ From: Rainer Krienke <krienke@uni-koblenz.de> Sent: 30 March 2021 13:05:56 To: Eugen Block; ceph-users@ceph.io Subject: [ceph-users] Re: ceph Nautilus lost two disk over night everything hangs Hello, yes your assumptions are correct pxa-rbd ist the metadata pool for pxa-ec which uses a erasure coding 4+2 profile. In the last hours ceph repaired most of the damage. One inactive PG remained and in ceph health detail then told me: --------- HEALTH_WARN Reduced data availability: 1 pg inactive, 1 pg incomplete; 15 daemons have recently crashed; 150 slow ops, oldest one blocked for 26716 sec, daemons [osd.60,osd.67] have slow ops. PG_AVAILABILITY Reduced data availability: 1 pg inactive, 1 pg incomplete pg 36.15b is remapped+incomplete, acting [60,2147483647,23,96,2147483647,36] (reducing pool pxa-ec min_size from 5 may help; search ceph.com/docs for 'incomplete') RECENT_CRASH 15 daemons have recently crashed osd.90 crashed on host ceph6 at 2021-03-29 21:14:10.442314Z osd.67 crashed on host ceph5 at 2021-03-30 02:21:23.944205Z osd.67 crashed on host ceph5 at 2021-03-30 01:39:14.452610Z osd.90 crashed on host ceph6 at 2021-03-29 21:14:24.222223Z osd.67 crashed on host ceph5 at 2021-03-30 02:35:43.373845Z osd.67 crashed on host ceph5 at 2021-03-30 01:19:58.762393Z osd.67 crashed on host ceph5 at 2021-03-30 02:09:42.297941Z osd.67 crashed on host ceph5 at 2021-03-30 02:28:29.981528Z osd.67 crashed on host ceph5 at 2021-03-30 01:50:05.374278Z osd.90 crashed on host ceph6 at 2021-03-29 21:13:51.896849Z osd.67 crashed on host ceph5 at 2021-03-30 02:00:22.593745Z osd.67 crashed on host ceph5 at 2021-03-30 01:29:39.170134Z osd.90 crashed on host ceph6 at 2021-03-29 21:14:38.114768Z osd.67 crashed on host ceph5 at 2021-03-30 00:54:06.629808Z osd.67 crashed on host ceph5 at 2021-03-30 01:10:21.824447Z --------- All osds except for 67 and 90 are up and I followed the hint in health detail and lowered min_size from 5 to 4 for pxa-ec. Since then ceph is again repairing and in between some VMs in the attached proxmox cluster are working again. So I hope that after repairing all PGs are up, so that I can restart all VMs again. Thanks Rainer Am 30.03.21 um 11:41 schrieb Eugen Block:
Hi,
from what you've sent my conclusion about the stalled I/O would be indeed the min_size of the EC pool. There's only one PG reported as incomplete, I assume that is the EC pool, not the replicated pxa-rbd, right? Both pools are for rbd so I'm guessing the rbd headers are in pxa-rbd while the data is stored in pxa-ec, could you confirm that?
You could add 'ceph health detail' output to your question to see which PG is incomplete. I assume that both down OSDs are in the acting set of the inactive PG, and since the pool's min_size is 5 the I/O pauses. If you can't wait for recovery to finish and can't bring up at least one of those OSDs you could set the min_size of pxa-ec to 4, but if you do, be aware that one more disk failure could mean data loss! So think carefully about it (maybe you could instead speed up recovery?) and don't forget to increase min_size back to 5 when the recovery has finished, that's very important!
Regards, Eugen
Zitat von Rainer Krienke <krienke@uni-koblenz.de>:
Hello,
i run a ceph Nautilus cluster with 9 hosts and 144 OSDs. Last night we lost two disks, so two OSDs (67,90) are down. The two disks are on two different hosts. A third ODS on a third host repotrts slow ops. ceph is repairing at the moment.
Pools affected are eg these ones: pool 35 'pxa-rbd' replicated size 3 min_size 2 crush_rule 0 object_hash rjenkins pg_num 256 pgp_num 256 last_change 192082 lfor 0/27841/27845 flags hashpspool,selfmanaged_snaps stripe_width 0 pg_num_min 128 target_size_ratio 0.0001 application rbd
pool 36 'pxa-ec' erasure size 6 min_size 5 crush_rule 7 object_hash rjenkins pg_num 512 pgp_num 512 last_change 192177 lfor 0/172580/172578 flags hashpspool,ec_overwrites,selfmanaged_snaps stripe_width 16384 pg_num_min 512 target_size_ratio 0.15 application rbd
At the mmoment the proxmox-cluster using storage from the seperate ceph cluster hangs. The ppols with date are erasure coded with the following profile:
crush-device-class= crush-failure-domain=host crush-root=default jerasure-per-chunk-alignment=false k=4 m=2 plugin=jerasure technique=reed_sol_van w=8
What I do not understand is why access on the virtualization seem to block. Could that be related to min_size of the pools cause this behaviour? How can I find out if this is true or what else is causing the blocking behaviour seen?
This is the current status: health: HEALTH_WARN Reduced data availability: 1 pg inactive, 1 pg incomplete Degraded data redundancy: 42384/130014984 objects degraded (0.033%), 4 pgs degraded, 5 pgs undersized 15 daemons have recently crashed 150 slow ops, oldest one blocked for 15901 sec, daemons [osd.60,osd.67] have slow ops.
services: mon: 3 daemons, quorum ceph2,ceph5,ceph8 (age 4h) mgr: ceph2(active, since 7w), standbys: ceph5, ceph8, ceph-admin mds: cephfsrz:1 {0=ceph6=up:active} 2 up:standby osd: 144 osds: 142 up (since 4h), 142 in (since 5h); 6 remapped pgs
task status: scrub status: mds.ceph6: idle
data: pools: 15 pools, 2632 pgs objects: 21.70M objects, 80 TiB usage: 139 TiB used, 378 TiB / 517 TiB avail pgs: 0.038% pgs not active 42384/130014984 objects degraded (0.033%) 2623 active+clean 3 active+undersized+degraded+remapped+backfilling 3 active+clean+scrubbing+deep 1 active+undersized+degraded+remapped+backfill_wait 1 active+undersized+remapped+backfill_wait 1 remapped+incomplete
io: client: 2.2 MiB/s rd, 3.6 MiB/s wr, 8 op/s rd, 179 op/s wr recovery: 51 MiB/s, 12 objects/s
Thanks a lot Rainer -- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312 _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
-- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312 _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
Hello Frank, the option is actually set. On one of my monitors: # ceph daemon /var/run/ceph/ceph-mon.*.asok config show|grep osd_allow_recovery_below_min_size "osd_allow_recovery_below_min_size": "true", Thank you very much Rainer Am 30.03.21 um 13:20 schrieb Frank Schilder:
Hi, this is odd. The problem with recovery when sufficiently many but less than min_size shards are present should have been resolved with osd_allow_recovery_below_min_size=true. It is really dangerous to reduce min_size below k+1 and, in fact, should never be necessary for recovery. Can you check if this option is present and set to true? If it is not working as intended, a tracker ticker might be in order.
Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14
-- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312
Dear Rainer, hmm, maybe the option is ignored or not implemented properly. This option set to true should have the same effect as reducing min_size *except* that new writes will not go to non-redundant storage. When reducing min-size, a critically degraded PG will accept new writes, which is the danger of data-loss mentioned before and avoided if only recovery ops are allowed on such PGs. Can you open a tracker about your observation that reducing min-size was necessary and helped despite osd_allow_recovery_below_min_size=true? Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14 ________________________________________ From: Rainer Krienke <krienke@uni-koblenz.de> Sent: 30 March 2021 13:30:00 To: Frank Schilder; Eugen Block; ceph-users@ceph.io Subject: Re: [ceph-users] Re: ceph Nautilus lost two disk over night everything hangs Hello Frank, the option is actually set. On one of my monitors: # ceph daemon /var/run/ceph/ceph-mon.*.asok config show|grep osd_allow_recovery_below_min_size "osd_allow_recovery_below_min_size": "true", Thank you very much Rainer Am 30.03.21 um 13:20 schrieb Frank Schilder:
Hi, this is odd. The problem with recovery when sufficiently many but less than min_size shards are present should have been resolved with osd_allow_recovery_below_min_size=true. It is really dangerous to reduce min_size below k+1 and, in fact, should never be necessary for recovery. Can you check if this option is present and set to true? If it is not working as intended, a tracker ticker might be in order.
Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14
-- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312
Sorry about the flow of messages. I forgot to mention this. Looking at other replies, in particular, that the PG in question remained at 4 out of 6 OSDs until you reduced min_size might indicate that peering was blocked for some reason, but completed after the reduction. If this was the order of events, it seems like an important detail. It is true that recovery will need to wait until the PG has the missing OSDs assigned. If this assignment is somehow blocked by min-size>k, the flag osd_allow_recovery_below_min_size itself will have no effect. Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14 ________________________________________ From: Frank Schilder Sent: 30 March 2021 14:53:18 To: Rainer Krienke; Eugen Block; ceph-users@ceph.io Subject: Re: [ceph-users] Re: ceph Nautilus lost two disk over night everything hangs Dear Rainer, hmm, maybe the option is ignored or not implemented properly. This option set to true should have the same effect as reducing min_size *except* that new writes will not go to non-redundant storage. When reducing min-size, a critically degraded PG will accept new writes, which is the danger of data-loss mentioned before and avoided if only recovery ops are allowed on such PGs. Can you open a tracker about your observation that reducing min-size was necessary and helped despite osd_allow_recovery_below_min_size=true? Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14 ________________________________________ From: Rainer Krienke <krienke@uni-koblenz.de> Sent: 30 March 2021 13:30:00 To: Frank Schilder; Eugen Block; ceph-users@ceph.io Subject: Re: [ceph-users] Re: ceph Nautilus lost two disk over night everything hangs Hello Frank, the option is actually set. On one of my monitors: # ceph daemon /var/run/ceph/ceph-mon.*.asok config show|grep osd_allow_recovery_below_min_size "osd_allow_recovery_below_min_size": "true", Thank you very much Rainer Am 30.03.21 um 13:20 schrieb Frank Schilder:
Hi, this is odd. The problem with recovery when sufficiently many but less than min_size shards are present should have been resolved with osd_allow_recovery_below_min_size=true. It is really dangerous to reduce min_size below k+1 and, in fact, should never be necessary for recovery. Can you check if this option is present and set to true? If it is not working as intended, a tracker ticker might be in order.
Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14
-- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312
I thought that recovery below min_size for EC pools wasn't expected to work until Octopus. From the Octopus release notes: "Ceph will allow recovery below min_size for Erasure coded pools, wherever possible." Josh On Tue, Mar 30, 2021 at 6:53 AM Frank Schilder <frans@dtu.dk> wrote:
Dear Rainer,
hmm, maybe the option is ignored or not implemented properly. This option set to true should have the same effect as reducing min_size *except* that new writes will not go to non-redundant storage. When reducing min-size, a critically degraded PG will accept new writes, which is the danger of data-loss mentioned before and avoided if only recovery ops are allowed on such PGs.
Can you open a tracker about your observation that reducing min-size was necessary and helped despite osd_allow_recovery_below_min_size=true?
Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14
________________________________________ From: Rainer Krienke <krienke@uni-koblenz.de> Sent: 30 March 2021 13:30:00 To: Frank Schilder; Eugen Block; ceph-users@ceph.io Subject: Re: [ceph-users] Re: ceph Nautilus lost two disk over night everything hangs
Hello Frank,
the option is actually set. On one of my monitors:
# ceph daemon /var/run/ceph/ceph-mon.*.asok config show|grep osd_allow_recovery_below_min_size "osd_allow_recovery_below_min_size": "true",
Thank you very much Rainer
Am 30.03.21 um 13:20 schrieb Frank Schilder:
Hi, this is odd. The problem with recovery when sufficiently many but less than min_size shards are present should have been resolved with osd_allow_recovery_below_min_size=true. It is really dangerous to reduce min_size below k+1 and, in fact, should never be necessary for recovery. Can you check if this option is present and set to true? If it is not working as intended, a tracker ticker might be in order.
Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14
-- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312 _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
Ahh, right. I saw it fixed here https://tracker.ceph.com/issues/18749 a long time ago, but it seems the back-port never happened. Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14 ________________________________________ From: Josh Baergen <jbaergen@digitalocean.com> Sent: 30 March 2021 15:23:10 To: Frank Schilder Cc: Rainer Krienke; Eugen Block; ceph-users@ceph.io Subject: Re: [ceph-users] Re: ceph Nautilus lost two disk over night everything hangs I thought that recovery below min_size for EC pools wasn't expected to work until Octopus. From the Octopus release notes: "Ceph will allow recovery below min_size for Erasure coded pools, wherever possible." Josh On Tue, Mar 30, 2021 at 6:53 AM Frank Schilder <frans@dtu.dk<mailto:frans@dtu.dk>> wrote: Dear Rainer, hmm, maybe the option is ignored or not implemented properly. This option set to true should have the same effect as reducing min_size *except* that new writes will not go to non-redundant storage. When reducing min-size, a critically degraded PG will accept new writes, which is the danger of data-loss mentioned before and avoided if only recovery ops are allowed on such PGs. Can you open a tracker about your observation that reducing min-size was necessary and helped despite osd_allow_recovery_below_min_size=true? Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14 ________________________________________ From: Rainer Krienke <krienke@uni-koblenz.de<mailto:krienke@uni-koblenz.de>> Sent: 30 March 2021 13:30:00 To: Frank Schilder; Eugen Block; ceph-users@ceph.io<mailto:ceph-users@ceph.io> Subject: Re: [ceph-users] Re: ceph Nautilus lost two disk over night everything hangs Hello Frank, the option is actually set. On one of my monitors: # ceph daemon /var/run/ceph/ceph-mon.*.asok config show|grep osd_allow_recovery_below_min_size "osd_allow_recovery_below_min_size": "true", Thank you very much Rainer Am 30.03.21 um 13:20 schrieb Frank Schilder:
Hi, this is odd. The problem with recovery when sufficiently many but less than min_size shards are present should have been resolved with osd_allow_recovery_below_min_size=true. It is really dangerous to reduce min_size below k+1 and, in fact, should never be necessary for recovery. Can you check if this option is present and set to true? If it is not working as intended, a tracker ticker might be in order.
Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14
-- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312 _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io<mailto:ceph-users@ceph.io> To unsubscribe send an email to ceph-users-leave@ceph.io<mailto:ceph-users-leave@ceph.io>
Hi, On 30.03.21 13:05, Rainer Krienke wrote:
Hello,
yes your assumptions are correct pxa-rbd ist the metadata pool for pxa-ec which uses a erasure coding 4+2 profile.
In the last hours ceph repaired most of the damage. One inactive PG remained and in ceph health detail then told me:
--------- HEALTH_WARN Reduced data availability: 1 pg inactive, 1 pg incomplete; 15 daemons have recently crashed; 150 slow ops, oldest one blocked for 26716 sec, daemons [osd.60,osd.67] have slow ops. PG_AVAILABILITY Reduced data availability: 1 pg inactive, 1 pg incomplete pg 36.15b is remapped+incomplete, acting [60,2147483647,23,96,2147483647,36] (reducing pool pxa-ec min_size from 5 may help; search ceph.com/docs for 'incomplete')
*snipsnap* 2147483647 is (uint32)(-1), which mean no associated OSD. So this PG does not have six independent OSDs, and no backfilling is happening since there are no targets to backfill. You mentioned 9 hosts, so if you use a simple host based crush rule ceph should be able to find new OSDs for that PG. If you do not use standard crush rules please check that ceph is able to derive enough OSDs to satisfy the PG requirements (six different OSDs). The 'incomplete' part might be a problem. If just a chunk would be missing, the state should be undersized, not incomplete... Regards, Burkhard
Hello, in between ceph is runing again normally, except for the two osds that are down because of the failed disks. What really helped in my situation was to lower min_size from 5 (k+1) to 4 in my 4+2 erasure code setup. So I am also greatful for the programmer who put the helping hint in ceph health detail for this situation. Thanks very much to everyone who answered my request to help out. What is left now is to replace the disks and then bring the two osds up again. Have a nice day Rainer Am 30.03.21 um 13:32 schrieb Burkhard Linke:
Hi,
On 30.03.21 13:05, Rainer Krienke wrote:
Hello,
yes your assumptions are correct pxa-rbd ist the metadata pool for pxa-ec which uses a erasure coding 4+2 profile.
In the last hours ceph repaired most of the damage. One inactive PG remained and in ceph health detail then told me:
--------- HEALTH_WARN Reduced data availability: 1 pg inactive, 1 pg incomplete; 15 daemons have recently crashed; 150 slow ops, oldest one blocked for 26716 sec, daemons [osd.60,osd.67] have slow ops. PG_AVAILABILITY Reduced data availability: 1 pg inactive, 1 pg incomplete pg 36.15b is remapped+incomplete, acting [60,2147483647,23,96,2147483647,36] (reducing pool pxa-ec min_size from 5 may help; search ceph.com/docs for 'incomplete')
*snipsnap*
2147483647 is (uint32)(-1), which mean no associated OSD. So this PG does not have six independent OSDs, and no backfilling is happening since there are no targets to backfill.
You mentioned 9 hosts, so if you use a simple host based crush rule ceph should be able to find new OSDs for that PG. If you do not use standard crush rules please check that ceph is able to derive enough OSDs to satisfy the PG requirements (six different OSDs).
The 'incomplete' part might be a problem. If just a chunk would be missing, the state should be undersized, not incomplete...
Regards,
Burkhard
_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
-- Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse 1 56070 Koblenz, Web: http://www.uni-koblenz.de/~krienke, Tel: +49261287 1312 PGP: http://www.uni-koblenz.de/~krienke/mypgp.html, Fax: +49261287 1001312
participants (5)
-
Burkhard Linke
-
Eugen Block
-
Frank Schilder
-
Josh Baergen
-
Rainer Krienke