Bug #1095
run "rados bench 10 seq -p data" print "error during benchmark: -5"
Status:
Closed
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:
0%
Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
Hi ,
ceph 0.27 ,
run "rados bench 10 seq -p data",sometimes , print "error during benchmark: -5".
=============================================================
logs:
- rados bench 10 write -p data
Maintaining 16 concurrent writes of 4194304 bytes for at least 10 seconds.
sec Cur ops started finished avg MB/s cur MB/s last lat avg lat
0 0 0 0 0 0 - 0
1 16 24 8 31.9944 32 0.989244 0.731578
2 16 36 20 39.9943 48 1.87479 0.911121
3 16 56 40 53.3266 80 0.986187 0.984835
4 16 74 58 57.9931 72 0.680331 0.931541
5 16 89 73 58.3932 60 0.445319 0.945591
6 16 106 90 59.9931 68 1.2443 0.986629
7 16 120 104 59.4219 56 1.34696 0.979426
8 16 136 120 59.9933 64 0.909194 0.977059
9 16 155 139 61.771 76 0.536061 0.97395
10 16 172 156 62.3932 68 0.562165 0.968041
11 10 173 163 59.2663 28 1.04106 0.980805
Total time run: 11.041616
Total writes made: 173
Write size: 4194304
Bandwidth (MB/sec): 62.672
Max latency: 2.4248
Min latency: 0.274832
- rados bench 10 seq -p data
sec Cur ops started finished avg MB/s cur MB/s last lat avg lat
0 0 0 0 0 0 - 0
1 16 36 20 79.986 80 0.634338 0.514692
2 16 55 39 77.9894 76 0.123429 0.598664
3 16 76 60 79.9899 84 0.395291 0.62393
4 16 97 81 80.9903 84 0.246663 0.665094
5 16 119 103 82.3906 88 1.01186 0.682504
6 16 139 123 81.991 80 0.835631 0.689855
7 16 161 145 82.8482 88 0.340252 0.694499
read got -2
error during benchmark: -5
error 5: Input/output error
History
#1 Updated by Sage Weil almost 13 years ago
- Status changed from New to Closed
Hi Jeff-
I think the problem here is just that the read phase is running out of data to read. Let the write phase run for longer.
#2 Updated by Greg Farnum almost 13 years ago
Actually, the write benchmark should record how many objects are left and the read benchmark isn't supposed to go past that number. I haven't looked at it in a while so I'm not sure how fragile the mechanics are, though...
Is this ever associated with other glitches or odd runs?