hi, 
I'm currently try to use ambari-log4j along to record our hive job workflow 
data in order to quantify the hive query performance. With the statistics of 
each job's performance parameters, we can find the bottleneck of the hive query 
e.g. data skew case and so on. But I find that when I launch a hive client, it 
will also load the hadoop log4j properties and thus create a ambari-log4j 
LoggingThread. The problem is when the database which we use to save workflow 
data is unavailable currently, we can't launch a hive client. I think the 
ambari-log4j is more like a plugin for hive and hadoop and it shouldn't affect 
the normal work of JobTracer or Hive. The attachment is a patch for this.
And another question workflowUpdateNumCompletedPS is incompatible with mysql.

-- 
chenchun

Attachment: ambari-log4j.patch.txt
Description: Binary data

Reply via email to