Affects Version/s: None
Today we can monitor the stream processing lag and latency outside of Nuxeo using stream.sh.
- can be used even when Nuxeo is down, only Kafka access is required
- don't impact Nuxeo perf
- more complex deployment, stream.sh being a standalone process, it needs to be run on multiple nodes to handle HA
- adding reporter implementation (JMX or Prometheus) requires specific implementation
In order to simplify the deployment, It will be better if these metrics are reported directly by Nuxeo, we will also gain the JMX and Prometheus reporting.
Because these metrics are global to the cluster we need to use the single producer pattern (a single thread reporting metrics in the entire cluster)
which is easy to do when using Kafka.