[NEMO-61] Fix lost execution metric collection problem

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 0.1
• Labels: pull-request-available
• Assignee: Jae Hyeon Park
• Reporter: Jeongyoon Eo
• Created: 10/May/18 07:46
• Updated: 29/Nov/18 09:41
• Resolved: 20/Jul/18 03:02

Description

We collect execution metrics in each Executor's MetricManagerWorker and send it to MetricManagerMaster, which looks like the following:

 

master.MetricManagerMaster: {"computationUnitId":"Stage-0-TaskGroup-409", "metricList":[
{"EndTime":1523534503827,"ContainerId":"Master","StartTime":1523534457564,"ToState":"ON_HOLD","ElapsedTime(ms)":46263,"FromState":"EXECUTING"}, {"EndTime":1523534504085,"InputReadTime(ms)":18660,"StartTime":1523534457869,"ReadBytes":0,"ElapsedTime(ms)":46216,"BoundedSourceReadTime(ms)":27393}]}

 

Currently, parts of TaskGroup execution metrics are lost occasionally. We need to fix this bug.

Issue Links

• blocks: NEMO-159 Nemo Web UI
• links to: GitHub Pull Request #60