In our test environment, we have a few flume 1.4 instances running. We are seeing the following when using the Http Json Monitoring:
1) Despite having multiple, distinct source-channel-sink combos, the counters for take attempt counts on all of our channels are set to the same exact non-zero number. Only the channel that actually gets written to should have this metric - we have validated that the others are not seeing data. 2) All success metrics across sources/channels/sinks show up as 0 even though data is clearly going into HDFS properly. 3) Over a seemingly random amount of time without any interference, this behavior eventually goes away without any changes. Not sure if anyone else uses these metrics, but if so have you seen this behavior?
