(Apache JIRA is down so continuing with email)
[~pallavi.rao], [~peeyushb], [~bvellanki], [~vrangan]
As it makes sense, I took a look at constructing the counters.txt path
mentioned below in the JMSMessageConsumer class.
{noformat}
6-06-29 02:46:11,574 INFO [main]
org.apache.falcon.workflow.WorkflowExecutionContext: counterFile -
hdfs://c6401.ambari.apache.org:8020/tmp/fs-backup/falcon/workflows/feed/sampleReplFeed/logs/job-2016-06-23-08-36/primaryCluster/counter.txt
{noformat}
From the OOZIE JMS message, I can get the entity name and entity type (decoding
the WF name), but the cluster names are not there -- in this case, the
replication WF runs on the destination cluster (backupCluster and so the
/tmp/fs-backup staging dir), but there is also primaryCluster in the path.
So, I think we need to rethink about where the counters file gets stored --
possibly in the STATS dir as pallavi points out or some other place that can be
derived independently.
Given that this is blocking 0.10 release, I would prefer to go ahead with Post
Processing sending the JMS message so that we can have this functionality work
and address the counters path in a separate JIRA.
What do people think?
