AnsweredAssumed Answered

JT Thinks all TT nodes are in default-rack

Question asked by thealy on Apr 11, 2013
Latest reply on Apr 12, 2013 by gera
Running M3, V2.1.1

While troubleshooting an unrelated issue and digging through the JobTracker log, I noticed that the JT seems to think all the TT nodes are in /default-rack/, while none of them actually are as confirmed on the Web console. Here are some JT log snippets. (Date removed to save space)

<code>
10:30:49,026 INFO org.apache.hadoop.mapred.JobTracker: Tasktracker tracker_hd21.xxx.yyy.zzz:localhost/127.0.0.1:34091 has sent a recovery heartbeat
10:30:49,028 INFO org.apache.hadoop.net.NetworkTopology: Adding a new node: /default-rack/hd21.xxx.yyy.zzz
10:30:49,028 INFO org.apache.hadoop.mapred.JobTracker: Adding tracker tracker_hd21.xxx.yyy.zzz:localhost/127.0.0.1:34091 to host hd21.xxx.yyy.zzz
....
10:31:43,527 INFO org.apache.hadoop.mapred.JobInProgress: tip:task_201304111030_0001_m_000042 has split on node:/default-rack/hd12.xxx.yyy.zzz
10:31:43,527 INFO org.apache.hadoop.mapred.JobInProgress: tip:task_201304111030_0001_m_000042 has split on node:/default-rack/hd38.sec.bnl.local
</code>

It seems that this would impact the assignment of "rack local" jobs and I'm not sure what else. Is this cause for concern?

-Terry

Outcomes