Ceph

10/22/2023

10:18 AM RADOS Bug #59831: crash: void ECBackend::continue_recovery_op(ECBackend::RecoveryOp&, RecoveryMessages*): assert(pop.data.length() == sinfo.aligned_logical_offset_to_chunk_offset( after_progress.data_recovered_to - op.recovery_progress.data_recovered_to))

enabling osd_read_ec_check_for_errors do fix this (https://tracker.ceph.com/issues/55662). And gives this log:
deb... 玮文 胡

07:09 AM RADOS Bug #59831: crash: void ECBackend::continue_recovery_op(ECBackend::RecoveryOp&, RecoveryMessages*): assert(pop.data.length() == sinfo.aligned_logical_offset_to_chunk_offset( after_progress.data_recovered_to - op.recovery_progress.data_recovered_to))

More info about the pool
pool 19 'cephfs.cephfs.data_ec' erasure profile clay_profile size 10 min_size 9 crush_rul... 玮文 胡

06:34 AM RADOS Bug #59831: crash: void ECBackend::continue_recovery_op(ECBackend::RecoveryOp&, RecoveryMessages*): assert(pop.data.length() == sinfo.aligned_logical_offset_to_chunk_offset( after_progress.data_recovered_to - op.recovery_progress.data_recovered_to))

We are hitting this in our production server (v17.2.6). Multiple OSDs crash loop. Recovery cannot progress. What can ... 玮文 胡

12/01/2022

03:41 AM RADOS Bug #53806: unessesarily long laggy PG state

Radoslaw Zarzynski wrote:
> OK, Aishwarya has found in testing that the @break@-related commit (https://github.com/c... 玮文 胡

06/29/2022

11:19 AM Linux kernel client Bug #53180: Attempt to access reserved inode number 0x101

OK, thanks. I forgot that [PATCH v2] because it is not listed under the same thread in https://lore.kernel.org/ceph-d... 玮文 胡

03:09 AM Linux kernel client Bug #53180: Attempt to access reserved inode number 0x101

I think the above patch is not yet pushed to the testing branch of https://github.com/ceph/ceph-client. Why is this i... 玮文 胡

05/25/2022

07:55 AM Ceph Revision 9558a6ad (ceph): qa: reproduce the crash when exporting unlinked dir

Signed-off-by: 胡玮文 <huww98@outlook.com> 玮文 胡

07:50 AM Ceph Revision fcdee486 (ceph): mds: fix crash when exporting unlinked dir

When fetch() an unlinked dir, we set fnode->version = 1, but leave
projected_version = 0. This will trigger the asser... 玮文 胡

05/23/2022

08:13 PM Ceph Revision 44cb4d9d (ceph): mon/LogMonitor: reopen log files on SIGHUP

Fixes: https://tracker.ceph.com/issues/55383
Signed-off-by: 胡玮文 <huww98@outlook.com>
(cherry picked from commit ee80f... 玮文 胡

05/03/2022

09:55 AM RADOS Bug #55383: monitor cluster logs(ceph.log) appear empty until rotated

Maybe we should fix the release note(https://docs.ceph.com/en/latest/releases/quincy/) first? The workaround there is... 玮文 胡