Hi, There are 2 distinct things indeed, I recommend you the following default value and tune from there:
max_queued_native_transport_requests=1024 This is the new default value after parch CASSANDRA-11363 - https://issues.apache.org/jira/secure/attachment/12818636/thread-queue-2.1.txt). You can always set it in the cassandra-env.sh file with "-Dcassandra.max_queued_native_transport_requests=1024" native_transport_max_threads: 128 Default is commented, it's better to control the incoming load explicitely, 128 is a commonly used value - https://github.com/apache/cassandra/blob/cassandra-3.11.0/conf/cassandra.yaml#L641-L645 Then use 'watch -d nodetool tpstats' or equivalent (monitoring...) on your nodes and if there is an increasing number of blocked 'Native-Transport-Requests', consider increasing the first value (max_queued_native_transport_requests=1024) if your machines can take more load. For the second parameter you can use something like 'netstats -tupawn | grep 9042 | grep ESTABLISHED | wc -l' or monitoring again. As long as this value is lower than 128, you are under the limit of connection. Note that in recent Native protocol, each connection can handle many simultaneous requests (128 before V3, 32768 for V3+). See the compatibility matrix for Native protocols http://docs.datastax.com/en/developer/java-driver/3.1/manual/native_protocol/ and the connection pooling I was mentioning above http://docs.datastax.com/en/developer/java-driver/3.1/manual/pooling/. C*heers, ----------------------- Alain Rodriguez - @arodream - al...@thelastpickle.com France The Last Pickle - Apache Cassandra Consulting http://www.thelastpickle.com 2017-08-09 4:49 GMT+01:00 Peng Xiao <2535...@qq.com>: > Dear All, > > any suggestion for optimal value for native_transport_max_threads? > as per https://issues.apache.org/jira/browse/CASSANDRA-11363,ma > x_queued_native_transport_requests=4096,how about native_transport_max_ > threads? > > Thanks, > Peng Xiao >
