Bug #18153
closedNo space cause osd crash (bluestore)
0%
Description
Ceph cluster will only one monitor and one osd, the disk size is 50GB
mon_osd_full_ratio is 95% mon_osd_nearfull_ratio is 85%,
create one image and keep on writing in it, when the usage of (ceph df) is near 70%,
ceph-osd crash because it could not alloc any resource (NO space)
I found from the code that the crash is an assert by designed, but I think the used
is not correct which shows by "ceph df" which causes the used could reach mon_osd_full_ratio
Files
Updated by Nathan Cutler over 7 years ago
- Related to Bug #16878: filestore: utilization ratio calculation does not take journal size into account added
Updated by Nathan Cutler over 7 years ago
Related to (possibly duplicate of) #16878
Updated by Nathan Cutler over 7 years ago
- Subject changed from No space cause osd crash to No space cause osd crash (bluestore)
Updated by Sage Weil over 7 years ago
- Status changed from New to Need More Info
Can you reproduce this with a bluestore log? (debug bluestore = 20)
Updated by zhiyuan wang over 7 years ago
- File ceph-osd.0.log ceph-osd.0.log added
I used version 11.0.2, and today I retested it with master, this issue is still there
Attached is the log with (debug bluestore = 20), due to the log size is very large with bluestore debug 20
I just get last 5000 lines, I guess it is enough.
Updated by Sage Weil about 7 years ago
- Status changed from Need More Info to Resolved
this fix has been backported to kraken; grab the latest kraken build