Graylog2
Datanode mem usage metrics incorrect
The memory usage metrics on my datanode don't match reality. They seem to ramp up for an hour and then drop back to where they started. In reality the memory usage stays mostly constant and has never reached more than 18% of the systems memory.
I think the rollup which happens every hour might cause the wrong calculation.
Your Environment
- Graylog Version: 6.0.0-rc2
This metric is showing the heap used/heap commited percentage. The dips are from where the gc kicks in. So, the graph is correct, but maybe not the most useful one for users. You could deduct if there is a memory leak if the baseline is rising over time. Do you have a suggestion for a better metric to monitor?
The dips are from where the gc kicks in
are you sure about that? The GC certainly runs not only hourly.
Or are you saying this is the usage percentage of the process internal pool?