Thread leak in cinder-backup with ceph backend
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Fix Released
|
High
|
Ivan Kolodyazhny |
Bug Description
Found on MOS 9.2 snapshot #688
After creating a backup, half of threads created by cinder-backup does not terminate.
This cause huge number of existing threads after a while:
root@node-49:~# ps -T -p $(pgrep cinder-backup) | wc -l
30966
Steps to reproduce:
1. Deploy environment with Ceph RBD as Cinder backend
2. Create an image: cinder create --name test-image 1
3. Note number of threads for cinder-backup: ps -T -p $(pgrep cinder-backup) | wc -l
4. Create a backup for the image: cinder create-backup test-image
5. After backup creation completes note number of threads again
Expected result: number of threads before and after creating backup is the same
Actual result: number of theads after creating backup is bigger (25 more)
The issue is not reproduced on devstack, so it seems to be already fixed in upstream
Changed in mos: | |
status: | New → Confirmed |
tags: | added: on-verification |
Changed in mos: | |
status: | Confirmed → In Progress |
Fix proposed to branch: 9.0/mitaka /review. fuel-infra. org/29654
Change author: cheng <email address hidden>
Review: https:/