Details
-
Bug
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
-
None
Description
'RuntimeMaster::flushMetrics()' sends requests to workers to flush all collected metrics to master. After then it saves all metrics to a file and optionally to DB.
The problem is that it does not check whether workers completely send metrics and master receives those metrics.
It's possible to miss some late metrics.
We need to fix this problem for correct measurement of system status, which is important for recent issues like Nemo-334.