How much memory is it use per one cardinality of meter?

[jaehleeee]

How much memory is it use per one cardinality of meter?

[jonatan-ivanov]

I guess you meant to ask per Meter? If so, which Meter? The documentation contains memory footprint estimation, you can take a look at it, for example for Timer: Timer memory footprint estimation. But instead of guessing, you can measure this. Create X amount of Meters and check your heap size before/after taking GC into consideration:

1. Start your app, warm it up with some "traffic"
2. Force full GC and make sure GC was triggered (e.g.: using GC logs)
3. Record heap size (this will be your baseline)
4. Create X amount of Meters
5. Force full GC and make sure GC was triggered (e.g.: using GC logs)
6. Record heap size

The diff between the two heap size recordings is roughly the memory consumption of X amount of Meters.

You can also check these:

• An article I wrote about high cardinality and what you can do with it: https://develotters.com/posts/high-cardinality/
• HighCardinalityTagsDetector to detect which tag(s) has(have) high cardinality.
• I made some measurements here: #3047
  Please notice that these measurements happened on my machine (architecture and OS), using my JVM (vendor and version) using my approximation methods and it might not be applicable to you.
• I also have a project called Resourceater that is a web application that is capable of creating resources (including Micrometer Meters) on demand.

[jaehleeee]

@jonatan-ivanov
I already read other issues about OOM by high cardinality (including https://develotters.com/posts/high-cardinality/)
I want to solve high cardinality of my application so I ask about how much memory use.

I mean Counter Meter and I want to know there are documents that already estimated about it.
(Before I start to measure this.)

Or maximum size of any meter?