Actions
Bug #59395
opencrush-device-class not respected
Status:
New
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:
0%
Source:
Tags:
Backport:
Regression:
No
Severity:
2 - major
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
I have a pool called CephFS-Erasure-ByOSD-D3F2-Backups, it is set to use the EC profile of Erasure-ByOSD-D3F2-Backups, that profile is set to use the device class of hdd-backups. But after creating and adding data to the pool, it doesn't put any PGs on the hdd-backups class.
# ceph osd pool get CephFS-Erasure-ByOSD-D3F2-Backups erasure_code_profile erasure_code_profile: Erasure-ByOSD-D3F2-Backups # ceph osd erasure-code-profile get Erasure-ByOSD-D3F2-Backups crush-device-class=hdd-backups crush-failure-domain=osd crush-root=default jerasure-per-chunk-alignment=false k=3 m=2 plugin=jerasure technique=reed_sol_van w=8
# rados df POOL_NAME USED OBJECTS CLONES COPIES MISSING_ON_PRIMARY UNFOUND DEGRADED RD_OPS RD WR_OPS WR USED COMPR UNDER COMPR CephFS-Erasure-ByOSD-D3F2-Backups 6.5 GiB 1692 0 8460 0 0 0 5 5 KiB 4381 12 GiB 0 B 0 B
# ceph osd df ID CLASS WEIGHT REWEIGHT SIZE RAW USE DATA OMAP META AVAIL %USE VAR PGS STATUS 7 hdd-backups 2.72899 1.00000 2.7 TiB 1.7 GiB 14 MiB 6 KiB 1.8 GiB 2.7 TiB 0.06 0.00 0 up 8 hdd-backups 2.72899 1.00000 2.7 TiB 517 MiB 14 MiB 5.0 MiB 511 MiB 2.7 TiB 0.02 0 0 up 9 hdd-backups 2.72899 1.00000 2.7 TiB 2.6 GiB 14 MiB 23 MiB 2.6 GiB 2.7 TiB 0.09 0.00 0 up 10 hdd-backups 2.72899 1.00000 2.7 TiB 1.6 GiB 14 MiB 4.5 MiB 1.6 GiB 2.7 TiB 0.06 0 0 up 22 hdd-backups 3.63869 1.00000 3.6 TiB 15 MiB 14 MiB 1 KiB 8.1 MiB 3.6 TiB 0 0 0 up
All of the PGs are places on the "hdd" crush-device-class.
I have not done additional testing at this point. My next step might be to use a device class without a special character (so "hddbackups" rather than "hdd-backups").
Ceph version 17.2.5.
Updated by Brian Woods about 1 year ago
Changing the class from "hdd-backups" to "backups" works around the problem.
There is defiantly an issue in the name processing for rules.
Actions