[jira] [Created] (AMBARI-11257) YARN Service Check job and Oozie job don't run because of whitespace in yarn.scheduler.capacity.root.default-node-label-expression

["Alejandro Fernandez (JIRA)" <ji...@apache.org>]

Alejandro Fernandez created AMBARI-11257:
--------------------------------------------

             Summary: YARN Service Check job and Oozie job don't run because of whitespace in yarn.scheduler.capacity.root.default-node-label-expression
                 Key: AMBARI-11257
                 URL: https://issues.apache.org/jira/browse/AMBARI-11257
             Project: Ambari
          Issue Type: Bug
            Reporter: Alejandro Fernandez


STR:
Create a cluster with Ambari 2.1.0, at build 1638
Install HDP 2.3.0.0 at build 2030, with HDFS, MR, YARN, ZK, OOZIE
Run Yarn Service Check
Run Oozie Service Check

The Resource Manager UI shows the apps as waiting for AM to schedule them, and eventually they get killed.

{code}
2015-05-19 22:15:28,976 DEBUG capacity.CapacityScheduler (CapacityScheduler.java:allocateContainersToNode(1168)) - Trying to schedule on node: knox-ru-4.c.pramod-thangali.internal, available: <memory:2048, vCores:1>
2015-05-19 22:15:28,976 INFO  scheduler.SchedulerUtils (SchedulerUtils.java:hasPendingResourceRequest(386)) - ****** partition= pending=<memory:0, vCores:0>
2015-05-19 22:15:28,976 INFO  scheduler.SchedulerUtils (SchedulerUtils.java:hasPendingResourceRequest(378)) -  ******** return false
2015-05-19 22:15:28,976 DEBUG capacity.ParentQueue (ParentQueue.java:assignContainers(394)) - Skip this queue=root, because it doesn't need more resource, schedulingMode=RESPECT_PARTITION_EXCLUSIVITY node-partition=
2015-05-19 22:15:29,538 DEBUG capacity.CapacityScheduler (CapacityScheduler.java:nodeUpdate(966)) - nodeUpdate: knox-ru-8.c.pramod-thangali.internal:45454 clusterResources: <memory:10240, vCores:5>
2015-05-19 22:15:29,538 DEBUG capacity.CapacityScheduler (CapacityScheduler.java:nodeUpdate(1013)) - Node being looked for scheduling knox-ru-8.c.pramod-thangali.internal:45454 availableResource: <memory:2048, vCores:1>
{code}

For Oozie, it is able to launch the MR jobs, but of course, they timeout.

The problem is that the yarn.scheduler.capacity.root.default-node-label-expression property is set to " " instead of "", and this behavior changed in HDP 2.3



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)