It was the driver after all. The C# driver (and I'm guessing others) query this
table as part of their heartbeat for idle connections. We have a lot of
clients. This adds up.
I don't believe this is the cause of the increasing network traffic.
On Wednesday, April 6, 2016 2:22 PM, Sotirios Delimanolis
wrote:
Hey,
I'm investigating an issue where the network traffic on a Cassandra 2.1 node
increases over time, regardless of the load our clients are under.
I tried enabling TRACE logging for org.apache.cassandra.transport.Message and
got bombarded with logs like these
DEBUG [SharedPool-Worker-2] 2016-04-06 21:10:09,416 Message.java:437 -
Received: QUERY SELECT key from system.local, v=3DEBUG [SharedPool-Worker-2]
2016-04-06 21:10:09,416 Message.java:452 - Responding: ROWS [key(system,
local), org.apache.cassandra.db.marshal.UTF8Type] | local---, v=3
We're talking A LOT of these. What is querying this system table? Drivers? Some
CassandraDB component? What other logs can I enable on Cassandra that will give
me more hints about what's going over the network?
Note that our clients are mostly Thrift-based.
Thanks,Soto