Occasional MASSIVE cpu metrics??

[MaxFlanders]

Occasionally dockerbeat is throwing a log with cpu.totalUsage off of the charts. Like unrealistically high, representing more cpu than our memory has (knowing you get 100% per cpu).

Example:

Have you guys seen this before or know what could cause something like this??

Thanks!

[tsg]

This looks quite similar to something we had in Topbeat, see this PR that fixes it. The usage of unsigned integers caused small negative values to be transformed in huge positive values.

Looking at the dockerbeat code, the same might be happening here.

[MaxFlanders]

Any update on this issue?? This is literally the only thing preventing us from utilizing this tool in production unfortunately :(

[marminthibaut]

I'll be on it on Thursday 🐎

[marminthibaut]

@326TimesBetter I just opened the PR #116 -> Can you check if this version of dockerbeat resolve your issue? If you don't want to compile dockerbeat yourselft, here is a compiled version for linux x86_64 architecture.

Thanks!
dockerbeat.zip

[marminthibaut]

@326TimesBetter ??

I wouldlike to release 😀