Ceph 16.2.3 issues during upgrade from 15.2.10 with cephadm/lvm list
Hi, We are seeing the mgr attempt to apply our OSD spec on the various hosts, then block. When we investigate, we see the mgr has executed cephadm calls like so, which are blocking: root 1522444 0.0 0.0 102740 23216 ? S 17:32 0:00 \_ /usr/bin/python3 /var/lib/ceph/XXXXX/cephadm.30cb78bdbbafb384af862e1c2292b944f15942b586128e91262b43e91e11ae90 --image docker.io/ceph/ceph@sha256:694ba9cdcbe6cb7d25ab14b34113c42c2d1af18d4c79c7ba4d1f62cf43d145fe ceph-volume --fsid XXXXX -- lvm list --format json This occurs on all hosts in the cluster, following starting/restarting/failing over a manager. It's blocking an in-progress upgrade post-manager updates on one cluster, currently. Looking at the cephadm logs on the host(s) in question, we see the last entry appears to be truncated, like: 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.db_uuid": "1n2f5v-EEgO-1Kn6-hQd2-v5QF-AN9o-XPkL6b", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.encrypted": "0", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osd_fsid": "XXXX", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osd_id": "205", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osdspec_affinity": "osd_spec", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.type": "block", The previous entry looks like this: 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.db_uuid": "TMTPD5-MLqp-06O2-raqp-S8o5-TfRG-hbFmpu", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.encrypted": "0", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.osd_fsid": "XXXX", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.osd_id": "195", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.osdspec_affinity": "osd_spec", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.type": "block", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.vdo": "0" 2021-05-10 17:32:06,470 INFO /usr/bin/podman: }, 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "type": "block", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "vg_name": "ceph-ffd1a4a7-316c-4c85-acde-06459e26f2c4" 2021-05-10 17:32:06,470 INFO /usr/bin/podman: } 2021-05-10 17:32:06,470 INFO /usr/bin/podman: ], We'd like to get to the bottom of this, please let us know what other information we can provide. Thank you, David
I think I may have found the issue: https://tracker.ceph.com/issues/50526 It seems it may be fixed in: https://github.com/ceph/ceph/pull/41045 I hope this can be prioritized as an urgent fix as it's broken upgrades on clusters of a relatively normal size (14 nodes, 24x OSDs, 2x NVME for DB/WAL w/ 12 OSDs per NVME), even when new OSDs are not being deployed, as it still tries to apply the OSD specification. On Mon, May 10, 2021 at 4:03 PM David Orman <ormandj@corenode.com> wrote:
Hi,
We are seeing the mgr attempt to apply our OSD spec on the various hosts, then block. When we investigate, we see the mgr has executed cephadm calls like so, which are blocking:
root 1522444 0.0 0.0 102740 23216 ? S 17:32 0:00 \_ /usr/bin/python3 /var/lib/ceph/XXXXX/cephadm.30cb78bdbbafb384af862e1c2292b944f15942b586128e91262b43e91e11ae90 --image docker.io/ceph/ceph@sha256:694ba9cdcbe6cb7d25ab14b34113c42c2d1af18d4c79c7ba4d1f62cf43d145fe ceph-volume --fsid XXXXX -- lvm list --format json
This occurs on all hosts in the cluster, following starting/restarting/failing over a manager. It's blocking an in-progress upgrade post-manager updates on one cluster, currently.
Looking at the cephadm logs on the host(s) in question, we see the last entry appears to be truncated, like:
2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.db_uuid": "1n2f5v-EEgO-1Kn6-hQd2-v5QF-AN9o-XPkL6b", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.encrypted": "0", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osd_fsid": "XXXX", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osd_id": "205", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osdspec_affinity": "osd_spec", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.type": "block",
The previous entry looks like this:
2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.db_uuid": "TMTPD5-MLqp-06O2-raqp-S8o5-TfRG-hbFmpu", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.encrypted": "0", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.osd_fsid": "XXXX", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.osd_id": "195", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.osdspec_affinity": "osd_spec", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.type": "block", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.vdo": "0" 2021-05-10 17:32:06,470 INFO /usr/bin/podman: }, 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "type": "block", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "vg_name": "ceph-ffd1a4a7-316c-4c85-acde-06459e26f2c4" 2021-05-10 17:32:06,470 INFO /usr/bin/podman: } 2021-05-10 17:32:06,470 INFO /usr/bin/podman: ],
We'd like to get to the bottom of this, please let us know what other information we can provide.
Thank you, David
The root cause is a bug in conmon. If you can upgrade to >= 2.0.26 this will also fix the problem. What version are you using? The kubic repos currently have 2.0.27. See https://build.opensuse.org/project/show/devel:kubic:libcontainers:stable We'll make sure the next release has the verbosity workaround! sage On Mon, May 10, 2021 at 5:47 PM David Orman <ormandj@corenode.com> wrote:
I think I may have found the issue:
https://tracker.ceph.com/issues/50526 It seems it may be fixed in: https://github.com/ceph/ceph/pull/41045
I hope this can be prioritized as an urgent fix as it's broken upgrades on clusters of a relatively normal size (14 nodes, 24x OSDs, 2x NVME for DB/WAL w/ 12 OSDs per NVME), even when new OSDs are not being deployed, as it still tries to apply the OSD specification.
On Mon, May 10, 2021 at 4:03 PM David Orman <ormandj@corenode.com> wrote:
Hi,
We are seeing the mgr attempt to apply our OSD spec on the various hosts, then block. When we investigate, we see the mgr has executed cephadm calls like so, which are blocking:
root 1522444 0.0 0.0 102740 23216 ? S 17:32 0:00 \_ /usr/bin/python3 /var/lib/ceph/XXXXX/cephadm.30cb78bdbbafb384af862e1c2292b944f15942b586128e91262b43e91e11ae90 --image docker.io/ceph/ceph@sha256:694ba9cdcbe6cb7d25ab14b34113c42c2d1af18d4c79c7ba4d1f62cf43d145fe ceph-volume --fsid XXXXX -- lvm list --format json
This occurs on all hosts in the cluster, following starting/restarting/failing over a manager. It's blocking an in-progress upgrade post-manager updates on one cluster, currently.
Looking at the cephadm logs on the host(s) in question, we see the last entry appears to be truncated, like:
2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.db_uuid": "1n2f5v-EEgO-1Kn6-hQd2-v5QF-AN9o-XPkL6b", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.encrypted": "0", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osd_fsid": "XXXX", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osd_id": "205", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osdspec_affinity": "osd_spec", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.type": "block",
The previous entry looks like this:
2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.db_uuid": "TMTPD5-MLqp-06O2-raqp-S8o5-TfRG-hbFmpu", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.encrypted": "0", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.osd_fsid": "XXXX", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.osd_id": "195", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.osdspec_affinity": "osd_spec", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.type": "block", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.vdo": "0" 2021-05-10 17:32:06,470 INFO /usr/bin/podman: }, 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "type": "block", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "vg_name": "ceph-ffd1a4a7-316c-4c85-acde-06459e26f2c4" 2021-05-10 17:32:06,470 INFO /usr/bin/podman: } 2021-05-10 17:32:06,470 INFO /usr/bin/podman: ],
We'd like to get to the bottom of this, please let us know what other information we can provide.
Thank you, David
_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
Hi Sage, We've got 2.0.27 installed. I restarted all the manager pods, just in case, and I have the same behavior afterwards. David On Mon, May 10, 2021 at 6:53 PM Sage Weil <sage@newdream.net> wrote:
The root cause is a bug in conmon. If you can upgrade to >= 2.0.26 this will also fix the problem. What version are you using? The kubic repos currently have 2.0.27. See https://build.opensuse.org/project/show/devel:kubic:libcontainers:stable
We'll make sure the next release has the verbosity workaround!
sage
On Mon, May 10, 2021 at 5:47 PM David Orman <ormandj@corenode.com> wrote:
I think I may have found the issue:
https://tracker.ceph.com/issues/50526 It seems it may be fixed in: https://github.com/ceph/ceph/pull/41045
I hope this can be prioritized as an urgent fix as it's broken upgrades on clusters of a relatively normal size (14 nodes, 24x OSDs, 2x NVME for DB/WAL w/ 12 OSDs per NVME), even when new OSDs are not being deployed, as it still tries to apply the OSD specification.
On Mon, May 10, 2021 at 4:03 PM David Orman <ormandj@corenode.com> wrote:
Hi,
We are seeing the mgr attempt to apply our OSD spec on the various hosts, then block. When we investigate, we see the mgr has executed cephadm calls like so, which are blocking:
root 1522444 0.0 0.0 102740 23216 ? S 17:32 0:00 \_ /usr/bin/python3 /var/lib/ceph/XXXXX/cephadm.30cb78bdbbafb384af862e1c2292b944f15942b586128e91262b43e91e11ae90 --image docker.io/ceph/ceph@sha256:694ba9cdcbe6cb7d25ab14b34113c42c2d1af18d4c79c7ba4d1f62cf43d145fe ceph-volume --fsid XXXXX -- lvm list --format json
This occurs on all hosts in the cluster, following starting/restarting/failing over a manager. It's blocking an in-progress upgrade post-manager updates on one cluster, currently.
Looking at the cephadm logs on the host(s) in question, we see the last entry appears to be truncated, like:
2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.db_uuid": "1n2f5v-EEgO-1Kn6-hQd2-v5QF-AN9o-XPkL6b", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.encrypted": "0", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osd_fsid": "XXXX", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osd_id": "205", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osdspec_affinity": "osd_spec", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.type": "block",
The previous entry looks like this:
2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.db_uuid": "TMTPD5-MLqp-06O2-raqp-S8o5-TfRG-hbFmpu", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.encrypted": "0", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.osd_fsid": "XXXX", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.osd_id": "195", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.osdspec_affinity": "osd_spec", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.type": "block", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.vdo": "0" 2021-05-10 17:32:06,470 INFO /usr/bin/podman: }, 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "type": "block", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "vg_name": "ceph-ffd1a4a7-316c-4c85-acde-06459e26f2c4" 2021-05-10 17:32:06,470 INFO /usr/bin/podman: } 2021-05-10 17:32:06,470 INFO /usr/bin/podman: ],
We'd like to get to the bottom of this, please let us know what other information we can provide.
Thank you, David
_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
We've created a PR to fix the root cause of this issue: https://github.com/alfredodeza/remoto/pull/63 Thank you, David On Mon, May 10, 2021 at 7:29 PM David Orman <ormandj@corenode.com> wrote:
Hi Sage,
We've got 2.0.27 installed. I restarted all the manager pods, just in case, and I have the same behavior afterwards.
David
On Mon, May 10, 2021 at 6:53 PM Sage Weil <sage@newdream.net> wrote:
The root cause is a bug in conmon. If you can upgrade to >= 2.0.26 this will also fix the problem. What version are you using? The kubic repos currently have 2.0.27. See https://build.opensuse.org/project/show/devel:kubic:libcontainers:stable
We'll make sure the next release has the verbosity workaround!
sage
On Mon, May 10, 2021 at 5:47 PM David Orman <ormandj@corenode.com> wrote:
I think I may have found the issue:
https://tracker.ceph.com/issues/50526 It seems it may be fixed in: https://github.com/ceph/ceph/pull/41045
I hope this can be prioritized as an urgent fix as it's broken upgrades on clusters of a relatively normal size (14 nodes, 24x OSDs, 2x NVME for DB/WAL w/ 12 OSDs per NVME), even when new OSDs are not being deployed, as it still tries to apply the OSD specification.
On Mon, May 10, 2021 at 4:03 PM David Orman <ormandj@corenode.com> wrote:
Hi,
We are seeing the mgr attempt to apply our OSD spec on the various hosts, then block. When we investigate, we see the mgr has executed cephadm calls like so, which are blocking:
root 1522444 0.0 0.0 102740 23216 ? S 17:32 0:00 \_ /usr/bin/python3 /var/lib/ceph/XXXXX/cephadm.30cb78bdbbafb384af862e1c2292b944f15942b586128e91262b43e91e11ae90 --image docker.io/ceph/ceph@sha256:694ba9cdcbe6cb7d25ab14b34113c42c2d1af18d4c79c7ba4d1f62cf43d145fe ceph-volume --fsid XXXXX -- lvm list --format json
This occurs on all hosts in the cluster, following starting/restarting/failing over a manager. It's blocking an in-progress upgrade post-manager updates on one cluster, currently.
Looking at the cephadm logs on the host(s) in question, we see the last entry appears to be truncated, like:
2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.db_uuid": "1n2f5v-EEgO-1Kn6-hQd2-v5QF-AN9o-XPkL6b", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.encrypted": "0", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osd_fsid": "XXXX", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osd_id": "205", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.osdspec_affinity": "osd_spec", 2021-05-10 17:32:06,471 INFO /usr/bin/podman: "ceph.type": "block",
The previous entry looks like this:
2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.db_uuid": "TMTPD5-MLqp-06O2-raqp-S8o5-TfRG-hbFmpu", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.encrypted": "0", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.osd_fsid": "XXXX", 2021-05-10 17:32:06,469 INFO /usr/bin/podman: "ceph.osd_id": "195", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.osdspec_affinity": "osd_spec", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.type": "block", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "ceph.vdo": "0" 2021-05-10 17:32:06,470 INFO /usr/bin/podman: }, 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "type": "block", 2021-05-10 17:32:06,470 INFO /usr/bin/podman: "vg_name": "ceph-ffd1a4a7-316c-4c85-acde-06459e26f2c4" 2021-05-10 17:32:06,470 INFO /usr/bin/podman: } 2021-05-10 17:32:06,470 INFO /usr/bin/podman: ],
We'd like to get to the bottom of this, please let us know what other information we can provide.
Thank you, David
_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
participants (2)
-
David Orman
-
Sage Weil