Actions
Bug #22963
closedradosgw-admin usage show loops indefinitly - again
% Done:
0%
Source:
Tags:
rgw, usage
Backport:
luminous
Regression:
No
Severity:
2 - major
Reviewed:
Affected Versions:
ceph-qa-suite:
rgw
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
After updating our cluster from jewel to luminous, radosgw-admin usage show
loops. I thought #21196 should have closed the Bug.
EG:radosgw-admin usage show --uid="DunkelDatenaustausch" --start-date=2018-01-01 --end-date=2018-02-01
{ "entries": [ { "user": "DunkelDatenaustausch", "buckets": [ { "bucket": "", "time": "2018-01-17 10:00:00.000000Z", "epoch": 1516183200, "owner": "DunkelDatenaustausch", "categories": [ { "category": "list_buckets", "bytes_sent": 344, "bytes_received": 0, "ops": 1, "successful_ops": 1 } ] }, { "bucket": "", "time": "2018-01-17 10:00:00.000000Z", "epoch": 1516183200, "owner": "DunkelDatenaustausch", "categories": [ { "category": "list_buckets", "bytes_sent": 344, "bytes_received": 0, "ops": 1, "successful_ops": 1 } ] }, { "bucket": "", "time": "2018-01-17 10:00:00.000000Z", "epoch": 1516183200, "owner": "DunkelDatenaustausch", "categories": [ { "category": "list_buckets", "bytes_sent": 344, "bytes_received": 0, "ops": 1, "successful_ops": 1 } ] }, [...]
Interestingly, the amount of data, bytes_sent and bytes_received seem to be correct and respect to complete time range. We wrote a little wrapper, that detects and breaks the loop. We still got the correct accounting, when we added the bytes afterwards.
For some accounts, we still get the original behaviour.
The cluster is rather old. It has been upgraded over years from bobtail, probably. I included the zonegroup config.
Files
Actions