CustomTThreadPoolServer should log TTransportException at DEBUG level
---------------------------------------------------------------------
Key: CASSANDRA-3142
URL: https://issues.apache.org/jira/browse/CASSANDRA-3142
Project: Cassandra
Issue Type: Bug
Reporter: Jim Ancona
Currently CustomTThreadPoolServer, like the Thrift TThreadPoolServer, silently
ignores TTransportException in its run() method. This is appropriate in most
cases because TTransportException occurs fairly often when client connections
die. However TTransportException is also thrown when TFramedTransport
encounters a frame that is larger than thrift_framed_transport_size_in_mb. In
that case, silently exiting the run loop leads to a SocketException on the
client side which can be both difficult to diagnose, in part because nothing is
logged by Cassandra, and high-impact, because the client may respond by marking
the server node down and retrying the too-large request on another node, where
it also fails. This process repeated leads to the entire cluster being marked
down (see https://github.com/rantav/hector/issues/212). I've filed two Thrift
issues (https://issues.apache.org/jira/browse/THRIFT-1323 and
https://issues.apache.org/jira/browse/THRIFT-1324), but in the meantime, I
suggest that CustomTThreadPoolServer log the exception at DEBUG level in order
to support easier troubleshooting.
I can submit a patch with the added log message.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
