Actions
Bug #52739
closedmsg/async/ProtocalV2: recv_stamp of a message is set to a wrong value
Status:
Resolved
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:
100%
Source:
Tags:
backport_processed
Backport:
octopus,pacific
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(RADOS):
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
ProtocalV2 sets the recv_stamp after the message is throttled and received completely.
This is wrong because it was supposed to be a timestamp indicating the messenger started to receive the message.
This will cause the Ceph Daemons won't be able to take account of the time that spent in messenger level
when they calculate the request latency.
And also some of the latency related perf counters will not include the time spent in messenger level, which will
bring troubles when diagnosing network-related performance issues.
Note that this is only for V2, V1 has correctly set this value .
Actions