It's difficult to distinguish between the logs produced by the Pilot process and the logs from the database subprocess.
For example:
kubectl --namespace test-cassandracluster-1512066728-5581 logs cass-cassandra-1512066728-5581-cassandra-ringnodes-1
...
I1130 18:36:33.090716 1 controller.go:64] Finished syncing pilot "test-cassandracluster-1512066728-5581/cass-cassandra-1512066728-5581-cassandra-ringnodes-0" (23.042µs)
I1130 18:36:33.090975 1 controller.go:31] processed work item
I1130 18:36:33.668826 1 controller.go:64] Finished syncing pilot "test-cassandracluster-1512066728-5581/cass-cassandra-1512066728-5581-cassandra-ringnodes-1" (14.238274ms)
I1130 18:36:33.669211 1 controller.go:31] processed work item
WARN 18:36:33 Seed gossip version is -2147483648; will not connect with that version
The last line is from Cassandra, but it's in a different format.
We could launch a goroutine to read from the stdout / stderr pipe of the process and feed it to glog.
Or we could reconsider a design where the database and the pilot run in separate containers, in which case there logs are kept separate anyway.
It's difficult to distinguish between the logs produced by the Pilot process and the logs from the database subprocess.
For example:
The last line is from Cassandra, but it's in a different format.
We could launch a goroutine to read from the stdout / stderr pipe of the process and feed it to glog.
Or we could reconsider a design where the database and the pilot run in separate containers, in which case there logs are kept separate anyway.
/kind bug