Bug #49879
closed
[pwl rwl] IO hang after a period of time I/O of different block sizes
Added by CONGMIN YIN about 3 years ago.
Updated about 2 years ago.
Description
For example, VM run on the rwl cache for a long time, because VM's IO block size is not a fixed value like 4k, but 512Bto tens of megabytes. After a long time, if the pmem cache is small, it will occur fragment issue due to the defect of pmdk allocator. https://github.com/ceph/ceph/pull/39915, this PR splited the IO into 1MB extent to reduce the probability of this problem, but the problem already exists. We plan to manage pmem space and allocation by ourselves in the future.
- Subject changed from librbd/cache/pwl: IO hang after a period of time rwl I/O of different block sizes to [pwl] IO hang after a period of time rwl I/O of different block sizes
- Status changed from New to In Progress
- Assignee set to CONGMIN YIN
- Project changed from Ceph to rbd
- Category deleted (
librbd)
- Status changed from In Progress to Fix Under Review
- Pull request ID set to 40493
- Status changed from Fix Under Review to Pending Backport
- Backport set to pacific
- Copied to Backport #50576: pacific: [pwl rwl] IO hang after a period of time I/O of different block sizes added
- Subject changed from [pwl] IO hang after a period of time rwl I/O of different block sizes to [pwl rwl] IO hang after a period of time I/O of different block sizes
- Status changed from Pending Backport to Resolved
While running with --resolve-parent, the script "backport-create-issue" noticed that all backports of this issue are in status "Resolved" or "Rejected".
Also available in: Atom
PDF