Riley Zimmerman

I've settled on a kairosdb.datastore.cassandra.write_buffer_max_size=25000 which is working relatively well for me now with KairosDB 1.1.3. The pauses are not as extreme as when I go with an even lower...

Thanks Brian! A few follow-up clarifications before closing out since I'm sure others could hit similar issues with the 1.1.3 retry code change. Steady state (almost all of the time)...

I've switched back to kairosdb.datastore.cassandra.write_buffer_max_size=500,000 in my dev env and I'm running more burst tests to narrow down what is going on with the failures. I've bumped up my KairosDB...

I've taken a few javacores of the locked up KairosDB JVM. Each javacore has the same 2 blocked threads. I have to kill KairosDB and restart it, but Cassandra does...

Hmmm, looks like those 2 threads are always blocked, even in a good environment. Something else must be locking up my KairosDB.

I'm thinking I'm stuck in the loop ` while (m_pendingMutations != null) ` line 342 of WriteBuffer.java https://github.com/kairosdb/kairosdb/blob/9469daa937056d84de121ff112a93947e44f960e/src/main/java/org/kairosdb/datastore/cassandra/WriteBuffer.java For the first 2 minutes I get messages "Error sending data to...

Thanks for the hector suggestion. I've specifically set kairosdb.datastore.cassandra.hector.retryDownedHosts=true and I'm playing around with the other settings. However, even with retryDownedHosts=true, once the connection to Cassandra has any issues it...

I've added a lot of tracing to my WriteBuffer.java. The WriteBuffer-X-# threads get an error trying to run m_pendingMutations.execute() inside the while loop of run() of WriteDataJob. The interesting thing...

It looks like the issue was **thrift_framed_transport_size_in_mb** in my cassandra.yaml. Default was 15 MB. Once I moved to 150 MB everything has been okay (so far), even with a kairosdb.datastore.cassandra.write_buffer_max_size...

Oops, I'm going back and editing out write_delay.