Back to bug 2093031

Who When What Removed Added
Vikhyat Umrao 2022-06-02 18:48:10 UTC Status NEW ASSIGNED
Red Hat One Jira (issues.redhat.com) 2022-06-02 18:56:19 UTC Link ID Red Hat Issue Tracker RHCEPH-4450
Vikhyat Umrao 2022-06-02 21:54:46 UTC Doc Type If docs needed, set a value Bug Fix
Vikhyat Umrao 2022-06-02 21:57:33 UTC Comment 0 updated
Vikhyat Umrao 2022-06-02 21:57:44 UTC Link ID Github ceph/ceph/pull/46252
Neha Ojha 2022-06-02 22:54:13 UTC Assignee nojha nmordech
Vikhyat Umrao 2022-06-02 23:02:18 UTC Summary pacific: osd/PGLog.cc: Trim duplicates by number of entries [RHCS 5.2] osd/PGLog.cc: Trim duplicates by number of entries
Veera Raghava Reddy 2022-06-05 16:41:56 UTC CC vereddy
Vikhyat Umrao 2022-06-06 18:06:00 UTC Blocks 2094069
Vikhyat Umrao 2022-06-08 17:21:51 UTC Flags needinfo?(nmordech)
CC nmordech
Vikhyat Umrao 2022-06-10 19:23:08 UTC Flags needinfo?(nmordech)
Vikhyat Umrao 2022-06-15 19:49:33 UTC Summary [RHCS 5.2] osd/PGLog.cc: Trim duplicates by number of entries [RHCS 5.2] dups.size logging + COT dups trim command + online dups trimming fix
Vikhyat Umrao 2022-06-15 19:51:40 UTC Assignee nmordech rzarzyns
Vikhyat Umrao 2022-06-15 19:52:12 UTC Link ID Red Hat Bugzilla 2093106
Vikhyat Umrao 2022-06-15 19:53:00 UTC Comment 0 updated
Vikhyat Umrao 2022-06-15 19:53:17 UTC Comment 0 updated
Vikhyat Umrao 2022-07-05 19:57:31 UTC Flags needinfo?(rzarzyns)
Flags needinfo?(nojha)
CC tserlin
Flags needinfo?(rzarzyns) needinfo?(nojha)
Vikhyat Umrao 2022-07-11 23:58:47 UTC Blocks 2101798
Vikhyat Umrao 2022-07-11 23:59:01 UTC Blocks 2101798
Dependent Products Red Hat OpenShift Data Foundation
Vikhyat Umrao 2022-07-11 23:59:41 UTC Blocks 2101798
Neha Ojha 2022-07-12 21:03:53 UTC Status ASSIGNED POST
errata-xmlrpc 2022-07-13 02:39:33 UTC Fixed In Version ceph-16.2.8-73.el8cp
Status POST MODIFIED
Status MODIFIED ON_QA
Vikhyat Umrao 2022-07-13 22:07:34 UTC Fixed In Version ceph-16.2.8-73.el8cp ceph-16.2.8-75.el8cp
Akash Raj 2022-07-29 06:51:12 UTC Blocks 2102272
Akash Raj 2022-07-29 06:52:01 UTC CC akraj
Flags needinfo?(rzarzyns)
Akash Raj 2022-08-01 08:24:23 UTC Flags needinfo?(rzarzyns)
Radoslaw Zarzynski 2022-08-03 11:19:30 UTC Doc Text Without this fix, trimming of PG log dups entries can get prevented during the low-level PG split operation which is used by the PG autoscaler with far higher frequency than by a human operator. Stalling the trimming of dups resulted in significant memory growth of PG log, leading to OSD crashes as it ran out of memory. Restarting an OSD did not solve the problem as the PG log is stored on disk and reloaded to RAM on startup.

With this fix, both off-line (using the `ceph-objectstore-tool` command) and on-line (within OSD) trimming is able to remove corrupted dups entries of a PG Log that jammed the on-line trimming machinery and were responsible for the memory growth. A debug improvement is implemented that prints the number of dups entries to the OSD's log to help future investigations.
Pawan 2022-08-04 10:10:00 UTC Status ON_QA VERIFIED
Radoslaw Zarzynski 2022-08-08 08:43:54 UTC Flags needinfo?(rzarzyns) needinfo?(rzarzyns)
Akash Raj 2022-08-09 07:42:17 UTC Doc Text Without this fix, trimming of PG log dups entries can get prevented during the low-level PG split operation which is used by the PG autoscaler with far higher frequency than by a human operator. Stalling the trimming of dups resulted in significant memory growth of PG log, leading to OSD crashes as it ran out of memory. Restarting an OSD did not solve the problem as the PG log is stored on disk and reloaded to RAM on startup.

With this fix, both off-line (using the `ceph-objectstore-tool` command) and on-line (within OSD) trimming is able to remove corrupted dups entries of a PG Log that jammed the on-line trimming machinery and were responsible for the memory growth. A debug improvement is implemented that prints the number of dups entries to the OSD's log to help future investigations.
Previously, trimming of PG log dups entries could be prevented during the low-level PG split operation, which is used by the PG autoscaler with far higher frequency than by a human operator. Stalling the trimming of dups resulted in significant memory growth of PG log, leading to OSD crashes as it ran out of memory. Restarting an OSD did not solve the problem as the PG log is stored on disk and reloaded to RAM on startup.

With this fix, both off-line (using the `ceph-objectstore-tool` command) and on-line (within OSD) trimming is able to remove corrupted dups entries of a PG Log that jammed the on-line trimming machinery and were responsible for the memory growth. A debug improvement is implemented that prints the number of dups entries to the OSD's log to help future investigations.
Akash Raj 2022-08-09 07:44:33 UTC Doc Text Previously, trimming of PG log dups entries could be prevented during the low-level PG split operation, which is used by the PG autoscaler with far higher frequency than by a human operator. Stalling the trimming of dups resulted in significant memory growth of PG log, leading to OSD crashes as it ran out of memory. Restarting an OSD did not solve the problem as the PG log is stored on disk and reloaded to RAM on startup.

With this fix, both off-line (using the `ceph-objectstore-tool` command) and on-line (within OSD) trimming is able to remove corrupted dups entries of a PG Log that jammed the on-line trimming machinery and were responsible for the memory growth. A debug improvement is implemented that prints the number of dups entries to the OSD's log to help future investigations.
.Corrupted dups entries of a PG Log can be removed by off-line and on-line trimming

Previously, trimming of PG log dups entries could be prevented during the low-level PG split operation, which is used by the PG autoscaler with far higher frequency than by a human operator. Stalling the trimming of dups resulted in significant memory growth of PG log, leading to OSD crashes as it ran out of memory. Restarting an OSD did not solve the problem as the PG log is stored on disk and reloaded to RAM on startup.

With this fix, both off-line (using the `ceph-objectstore-tool` command) and on-line (within OSD) trimming is able to remove corrupted dups entries of a PG Log that jammed the on-line trimming machinery and were responsible for the memory growth. A debug improvement is implemented that prints the number of dups entries to the OSD's log to help future investigations.
errata-xmlrpc 2022-08-09 10:00:26 UTC Status VERIFIED RELEASE_PENDING
errata-xmlrpc 2022-08-09 17:39:07 UTC Status RELEASE_PENDING CLOSED
Resolution --- ERRATA
Last Closed 2022-08-09 17:39:07 UTC
errata-xmlrpc 2022-08-09 17:39:39 UTC Link ID Red Hat Product Errata RHSA-2022:5997
Vikhyat Umrao 2022-08-19 16:02:48 UTC Blocks 2119853
Vikhyat Umrao 2022-08-29 23:29:25 UTC CC linuxkidd
Link ID Red Hat Knowledge Base (Solution) 6963402
Dependent Products Red Hat OpenShift Data Foundation

Back to bug 2093031