Selvaganesan Govindarajan created TRAFODION-2596:
----------------------------------------------------
Summary: Improve the log4j and log4cxx infrastructure in Trafodion
Key: TRAFODION-2596
URL: https://issues.apache.org/jira/browse/TRAFODION-2596
Project: Apache Trafodion
Issue Type: Improvement
Reporter: Selvaganesan Govindarajan
Currently SQL logs messages in different files identified by the master node
and pid in its name. This creates many log files in $TRAF_HOME/log directory
and makes it unmanageable.
SQL primarily use log4cxx infrastructure for logging error messages and error
events. I believe error messages are written only when the master process reads
the diagnostics area and it is not written when the error originates. But the
error events(SQLMXLogging) might be written by any process.
In addition, all other logs on the java side uses a single log file even when
it is written from multiple processes in our environment. Eg
trafodion.hdfs.log trafodion.dtm.log. Even the C++ logging writes from multiple
processes. The amount of log entries written by SQL from C++ side is way less
than the any other logging within our environment. Definitely, they pale in
comparison with hbase, hive and other Hadoop processes.
So, I think it will be neat solution to log the entries from all SQL processes
into one log file per node
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
