Back to bug 2208600

Who When What Removed Added
Red Hat One Jira (issues.redhat.com) 2023-05-19 17:29:39 UTC Link ID Red Hat Issue Tracker RHCEPH-6703
Sridhar Seshasayee 2023-05-19 17:31:41 UTC Severity unspecified medium
Vikhyat Umrao 2023-05-19 19:14:53 UTC Flags needinfo?(asriram)
CC asriram, vumrao
Target Release 6.1z1 6.1
Ranjini M N 2023-05-23 12:43:58 UTC Assignee ceph-docs rmandyam
Status NEW ASSIGNED
CC rmandyam
Docs Contact asriram akraj
Ranjini M N 2023-05-24 08:15:20 UTC QA Contact msaini skanta
CC skanta
Assignee rmandyam ceph-docs
Docs Contact akraj
Sub Component Configurations Guide Administration Guide
QA Contact skanta msaini
QA Contact msaini skanta
Assignee ceph-docs rmandyam
Ranjini M N 2023-05-25 09:10:21 UTC Flags needinfo?(asriram) needinfo?(akraj)
CC akraj
Docs Contact akraj
Ranjini M N 2023-05-25 09:11:29 UTC Status ASSIGNED POST
Akash Raj 2023-05-25 11:26:54 UTC Flags needinfo?(akraj)
Ranjini M N 2023-05-26 06:32:17 UTC Status POST ON_QA
Veera Raghava Reddy 2023-05-26 11:04:07 UTC CC vereddy
Sridhar Seshasayee 2023-06-02 12:11:04 UTC Status ON_QA ASSIGNED
Flags needinfo?(sseshasa)
Flags needinfo?(sseshasa)
Akash Raj 2023-06-05 14:30:56 UTC CC msaini
Flags needinfo?(sseshasa) needinfo?(msaini)
Status ASSIGNED ON_QA
Manisha Saini 2023-06-05 18:55:30 UTC Flags needinfo?(msaini)
Rivka Pollack 2023-06-06 07:57:01 UTC Assignee rmandyam rpollack
CC rpollack
Sridhar Seshasayee 2023-06-06 08:37:35 UTC Flags needinfo?(sseshasa)
Akash Raj 2023-06-06 11:47:59 UTC Flags needinfo?(sseshasa)
Sridhar Seshasayee 2023-06-06 12:37:01 UTC Flags needinfo?(sseshasa)
Sridhar Seshasayee 2023-06-07 07:13:47 UTC Flags needinfo?(sseshasa)
Flags needinfo?(sseshasa)
Rivka Pollack 2023-06-07 14:13:04 UTC Blocks 2192813
Docs Contact akraj rpollack
Assignee rpollack akraj
Akash Raj 2023-06-08 07:33:38 UTC Flags needinfo?(sseshasa) needinfo?(skanta)
Akash Raj 2023-06-08 10:16:02 UTC Flags needinfo?(sseshasa)
Sridhar Seshasayee 2023-06-08 15:35:33 UTC Flags needinfo?(skanta)
Flags needinfo?(sseshasa)
Flags needinfo?(akraj)
Flags needinfo?(sseshasa)
Flags needinfo?(sseshasa)
Flags needinfo?(sseshasa)
Flags needinfo?(akraj)
Doc Type If docs needed, set a value Enhancement
Doc Text Feature:
Improve the Ceph mClock OSD Scheduler documentation according to bug fixes and enhancements made to the mClock scheduler in RHCS 6.1.

Reason:
Significant usability and design improvements were made to the mClock scheduler to address the slow backfill issues in RHCS 6.0.

Result:
Important documentation highlights are:
* The 'balanced' profile is set as the default mClock profile because it represents a compromise between prioritizing client IO or recovery IO. Users can then choose either the 'high_client_ops' profile to prioritize client IO or the 'high_recovery_ops' profile to prioritize recovery IO.
* QoS parameters like reservation and limit are now specified in terms of a fraction (range: 0.0 to 1.0) of the OSD's IOPS capacity.
* The cost parameters (osd_mclock_cost_per_io_usec_* and
osd_mclock_cost_per_byte_usec_*) have been removed. The cost of an operation is now determined using the random IOPS and maximum sequential bandwidth capability of the OSD's underlying device.
* Degraded object recovery is given higher priority when compared to misplaced object recovery because degraded objects present a data safety issue not present with objects that are merely misplaced. Therefore, backfilling operations with the 'balanced' and 'high_client_ops' mClock profiles may progress slower than what was seen with the 'WeightedPriorityQueue' (WPQ) scheduler.
* The QoS allocations in all the mClock profiles are optimized based on the above fixes and enhancements.
Sridhar Seshasayee 2023-06-09 03:51:06 UTC Flags needinfo?(sseshasa)
Flags needinfo?(sseshasa)
Akash Raj 2023-06-09 04:38:16 UTC Status ON_QA ASSIGNED
Flags needinfo?(sseshasa)
Akash Raj 2023-06-09 04:39:21 UTC Status ASSIGNED ON_QA
Flags needinfo?(skanta)
Sridhar Seshasayee 2023-06-09 05:21:37 UTC Flags needinfo?(sseshasa)
Akash Raj 2023-06-09 07:44:08 UTC Flags needinfo?(skanta)
Status ON_QA VERIFIED
Doc Text Feature:
Improve the Ceph mClock OSD Scheduler documentation according to bug fixes and enhancements made to the mClock scheduler in RHCS 6.1.

Reason:
Significant usability and design improvements were made to the mClock scheduler to address the slow backfill issues in RHCS 6.0.

Result:
Important documentation highlights are:
* The 'balanced' profile is set as the default mClock profile because it represents a compromise between prioritizing client IO or recovery IO. Users can then choose either the 'high_client_ops' profile to prioritize client IO or the 'high_recovery_ops' profile to prioritize recovery IO.
* QoS parameters like reservation and limit are now specified in terms of a fraction (range: 0.0 to 1.0) of the OSD's IOPS capacity.
* The cost parameters (osd_mclock_cost_per_io_usec_* and
osd_mclock_cost_per_byte_usec_*) have been removed. The cost of an operation is now determined using the random IOPS and maximum sequential bandwidth capability of the OSD's underlying device.
* Degraded object recovery is given higher priority when compared to misplaced object recovery because degraded objects present a data safety issue not present with objects that are merely misplaced. Therefore, backfilling operations with the 'balanced' and 'high_client_ops' mClock profiles may progress slower than what was seen with the 'WeightedPriorityQueue' (WPQ) scheduler.
* The QoS allocations in all the mClock profiles are optimized based on the above fixes and enhancements.
.Usability and design improvements are implemented to the mClock scheduler
With this release, significant usability and design improvements are implemented to the mClock scheduler to address the slow backfill issues.
- The `balanced` profile is set as the default mClock profile because it represents a compromise between prioritizing client IO or recovery IO. Users can then choose either the `high_client_ops` profile to prioritize client IO or the `high_recovery_ops` profile to prioritize recovery IO.
- QoS parameters like _reservation_ and _limit_ are now specified in terms of a fraction (`range`: `0.0` to `1.0`) of the OSD's IOPS capacity.
- The cost parameters - `osd_mclock_cost_per_io_usec_*` and `osd_mclock_cost_per_byte_usec_*` have been removed. The cost of an operation is now determined using the random IOPS and maximum sequential bandwidth capability of the OSD's underlying device.
- Degraded object recovery is given higher priority when compared to misplaced object recovery because degraded objects present a data safety issue not present with objects that are merely misplaced. Therefore, backfilling operations with the `balanced` and `high_client_ops` mClock profiles may progress slower than what was seen with the `WeightedPriorityQueue` (WPQ) scheduler.
- The QoS allocations in all the mClock profiles are optimized based on the above fixes and enhancements.
Akash Raj 2023-06-09 09:45:02 UTC Status VERIFIED RELEASE_PENDING
Ranjini M N 2023-06-21 06:59:51 UTC Resolution --- CURRENTRELEASE
Status RELEASE_PENDING CLOSED
Last Closed 2023-06-21 06:59:51 UTC

Back to bug 2208600