When having a cluster of process engines, they will. currently send the same metrics data they obtain from the database. This will make the data meaningless on ET side.

AT:

• Change the implementation to only send metrics that were created on the current engine (and collected in main memory)
• It is okay that metrics may be lost when the engine shuts down unnormally

Other solution options:

1. Build a mechanism that ensures only one reporter in a cluster sends the metrics
   • Very hard to get right, taking into account that engines can start and stop at any time
2. Use the reporter id for the metrics query, sending only metrics that were created with the id of the current engine
   • It is possible that two engines use the same reporter id. In that case, we would again count metrics multiple times. We decided it is better to have a value lower than the true metrics instead of a higher value.
3. Send information with the metrics that allows to de-duplicate the metrics on Kibana side
   • Examples:
     • Send a unique reporter id that is unique per reporter, so that on Kibana side we use the metrics from only one reporter. Problem: The reporter id does not remain stable over engine restarts.
     • Send a time window for which the metrics were collected. On Kibana side we use only one data point per time window. Problem: Time windows of multiple reporters will not be exactly the same but overlap. Then, it is impossible to properly de-duplicate the data.