I have had memory reclaim problems at my jsreport server for several years. The last time I digged into this problem, I "closed the case", after discussion with Jan. We thought that the memory in the graph would be automatically garbage collected when needed. But I don'think this works very good.
This has seldom been a problem since I'm queing all my report requests on a separate server, which feeds jsreport with jobs. Failed reports will eventually be rendered and the client probably won't notice the delay caused by the jsreport restart.
When jsreport is fed with very large reports (json dataset 20-40 MB) this gets troublesome.
This memory chart shows the memory consumption for tre consecutive large reports. The third report crashes my Fargate container.
I'm running the official 2.11 Docker image.
Strategy is "in-process".
The machine is a 2vCpu 5GB Fargate container.
How should I investigate this further?
I'm assuming that this is some kind of out-of-memory crash, but I don't know since the crashed machine does lot leave any evidence after the crash.