Bug #19487
closed
"GLOBAL %RAW USED" of "ceph df" is not consistent with check_full_status
Added by Pan Liu about 7 years ago.
Updated over 6 years ago.
Category:
Administration/Usability
Description
1) Use vstart.sh to create a cluster, with option: osd failsafe ful ratio = .46
2) Input "ceph df":
GLOBAL:
SIZE AVAIL RAW USED %RAW USED
439G 217G 199G 45.45
3) rbd create image --size 128
4) ceph -w:
2017-04-04 20:57:40:894659 [ERR] OSD full dropping all updates 51% full
2017-04-04 20:57:44:364939 [ERR] pgmap v31: 24 pgs: 24 active+clean: 2068 bytes data, 199 GB used, 217 GB / 439 GB avail
So we could see ceph -w displays it is full(51%), but ceph df only 45.45%.
- Status changed from New to Fix Under Review
Let's say I set osd failsafe ful ratio = .90. Below I made up these numbers to show how
these percentages won't necessarily match anyway.
dzafman$ bin/ceph df
- DEVELOPER MODE: setting PATH, PYTHONPATH and LD_LIBRARY_PATH ***
GLOBAL:
SIZE AVAIL RAW USED %RAW USED
400G 69G 331G 82.75
[~/ceph/build] (wip-15912-followon)
dzafman$ df -h
Filesystem Size Used Avail Use% Mounted on
/dev/sdb1 100G 80G 20G 60% /var/lib/ceph-0
/dev/sdc1 100G 80G 20G 60% /var/lib/ceph-1
/dev/sdd1 100G 80G 20G 60% /var/lib/ceph-2
/dev/sde1 100G 91G 9G 91% /var/lib/ceph-3
osd.3 will report failsafe full but it won't match the global used percentage.
The only time it would come into play would be if every OSD had the same usage and it was 91% across the board on all OSDs so the "ceph df"
should also be 91%.
- Project changed from Ceph to RADOS
- Category set to Administration/Usability
- Status changed from Fix Under Review to In Progress
- Assignee set to David Zafman
Based on PR comments we expect this to be fixed up by one of David's disk handling branches. Or did that one already merge?
- Status changed from In Progress to Closed
Reopen this if issue hasn't been fixed in the latest code with the understanding that each OSD has its own fullness determination.
Also available in: Atom
PDF