[
https://issues.apache.org/jira/browse/HADOOP-5848?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12709925#action_12709925
]
Edward Capriolo commented on HADOOP-5848:
-----------------------------------------
Looking at latest trunk I do see. TaskTrackerMetricsInst
{noformat}
private int numCompletedTasks = 0;
private int timedoutTasks = 0;
private int tasksFailedPing = 0;
{noformat}
{noformat}
metricsRecord.setMetric("maps_running", tt.mapTotal);
metricsRecord.setMetric("reduces_running", tt.reduceTotal);
{noformat}
I would like to suggest adding:
{noformat}
private int numStartedTasks = 0;
{noformat}
With numStartedTasks your can determine.
{noformat}
tasksrunning=started-completed-failed_timeout-failed_ping
{noformat}
> JMX Metrics For TaskTracker
> ---------------------------
>
> Key: HADOOP-5848
> URL: https://issues.apache.org/jira/browse/HADOOP-5848
> Project: Hadoop Core
> Issue Type: New Feature
> Components: metrics
> Reporter: Edward Capriolo
> Priority: Minor
>
> Create useful metrics for the TaskTracker Component.
> * TasksStarted: Counter
> * TasksFailed: Counter
> If graphed, having these metrics will help administrators determine
> graphically how task load is being spread. Also will help determine if
> certain nodes have external conditions that are causing a task to fail.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
