Large allocation samples has too many points

[polarnik]

Expected behavior

Large allocation samples panel will have a few parts:

• up to 10 MByte
• 10-20
• 20-50
• ...
• 1000+

Actual behavior

Dashboard 0.5.0 has one panel Large allocation samples for all large objects. As a result graph has info about all points, it doesn't work well, because the total amount of points is too high. The table has info about big objects, viewers will not stroll table rows, they will skip info about middle-size and small-size objects

Steps to reproduce the problem

JFR-Exporter Version

0.5.0

Java Version

Java 21

Docker image: maven:3.9.9-amazoncorretto-21-debian

OS Version

Debian

Docker image: maven:3.9.9-amazoncorretto-21-debian

InfluxDB Version

2.7.10

Docker image: influxdb:2.7.10

Grafana Version

10.4.7

Docker image: grafana/grafana:10.4.7

Grafana dashboard version

0.5.0

[polarnik]

We can split metrics by threads and stack-traces only

[PeterPaul-Perfana]

Hi @polarnik, interesting idea to split it up by size. Looks useful. We can define a separate "memory allocation details" dashboard. I am curious for the use cases.

As an aside: users can currently also filter on object size, so only the large ones are collected. When these location are "fixed" (assuming the goal is to get rid of big allocations where possible) the object size threshold can be lowered.