Ceph Community News (2022-01-17 to 2022-02-16)
Cephalocon 2022 Postponed
Due to the COVID-19 pandemic, Cephalocon 2022, originally scheduled for April 5-7 (April 6-8, Beijing Time), has been postponed. The new time is yet to be determined. The topics of the conference have been released and are listed below. For more details, see Cephalocon 2022 schedule.
Recently Merged PRs
Recently, PRs have mainly focused on bug fixing. The following describes notable changes:
- mgr: Disabled the pg recovery by default when OSD is in or out. You can manually enable it as required to reduce the impact on the service cluster. pr#44588
- osd: Added the dump_blocked_ops_count option to ceph daemon perf dump to quickly obtain the number of blocked ops, reducing the overhead caused by the dump_blocked_ops operation. pr#44780
- rgw: Added support for the conditional copy by the rgw s3 CopyObject interface. pr#44678
- rgw: Fixed the issue that a large amount of memory is used during the radosgw-admin bucket chown process. pr#44357
- rbd: Introduced the rxbounce option to krbd to solve the problem of CRC errors and performance deterioration when images are used as block devices of the Windows system. pr#44842
Recent Ceph Developer News
Each module of the Ceph community holds regular meetings to discuss and align the development progress. Meeting videos are uploaded to YouTube. The major meetings are as follows:
Meeting | Description | Frequency |
---|---|---|
Crimson SeaStore OSD Weekly Meeting | Crimson & SeaStore development | Weekly |
Ceph Orchestration Meeting | Ceph management module (mgr) development | Weekly |
Ceph DocUBetter Meeting | Document optimization | Biweekly |
Ceph Performance Meeting | Ceph performance optimization | Biweekly |
Ceph Developer Monthly | Ceph developers | Monthly |
Ceph Testing Meeting | Version verification and release | Monthly |
Ceph Science User Group Meeting | Ceph scientific computing | Irregularly |
Ceph Leadership Team Meeting | Ceph leadership team | Weekly |
Recently, the community focuses on the freeze test and verification of the Quincy version. The following topics are discussed at the meetings:
- In the Quincy version test, the read performance meets the expectation, but the write performance deteriorates in some scenarios. It can be determined that 4k min_alloc_size and bluestore allocator can improve the performance.
- As the omap scale increases, the omap_iterator efficiency causes a large number of slow_ops or even no response. The issue records the test results of the two compaction modes. If the compaction is manually triggered, the latency cannot be restored to the previous level. To resolve this issue, Rocksdb provides periodic and TTL compaction. After this function is enabled, the latency can be restored to the previous level.
- A large number of PRs are merged for the Ceph Benchmark Tool (CBT) with a focus on the control of memory resources during the large-scale OSD test and the multi-client concurrent test cases.