streaming is hard-coded to be single-threaded
---------------------------------------------
Key: CASSANDRA-3549
URL: https://issues.apache.org/jira/browse/CASSANDRA-3549
Project: Cassandra
Issue Type: Bug
Reporter: Peter Schuller
Assignee: Peter Schuller
Priority: Minor
It turns out, after we were scratching our heads as to why we weren't streaming
as fast as expected when we bootstrapped 24 additional nodes in a 24 node
cluster, that a node will only ever stream one outgoing file at a time. This is
because the MessagingService instantiates it's streamExecturor_ in such a way
that it gets a maximum size of 1.
This is particularly bad because there is no mechanism to ensure that
destination nodes stream "evenly" from availabe source nodes, so some source
nodes were dogpiled.
Meanwhile, the streams that were active were bottlenecking (in our case) on the
destination side due to writes. In the end the aggregate throughput
bootstrapping in the cluster was far below intended.
Will submit a patch once we've confirmed it seems safe and that there's no
concurrency issue lurking.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
