Workarounds for hyperkit CPU problems?

I’m running into the dreaded com.docker.hyperkit cpu usage problems with docker for mac on macos sierra and latest docker for mac. Specifically I’m trying to generate a jar and it’s taking over 8 minutes to generate a single uberjar! Same jar takes 2 seconds on linux!

It’s basically makes docker un-usable. Any workarounds?

Is there a beta version I can run?

Would it be possible to post a reproduction case at all? That sounds like some pathological behavior and we’d really like to understand what exactly your use case is doing so we can improve the software. There is a beta channel available from https://docs.docker.com/docker-for-mac/ but I’m uncertain if it will satisfy you. The best way forward is with a reproduction and maybe there is a straightforward bug to fix to improve performance of your use case in the coming betas.

Thanks!

It looks like the problem is already known. https://github.com/docker/for-mac/issues/668

It seems to be related to read/write on shared volumes.

I installed the beta and made the fixes suggested in the github issue thread, but they had no effect.

@mikeball, the problem is not already known – you are experiencing several orders of magnitude worse performance than expected (even with the known performance problems). We’d really appreciate a reproduction case as you seem to be hitting an unknown performance problem. Could you share your use case in a way that enables us to reproduce it locally?

Here’s a repository that should allow you to replicate.

1 Like

I am experiencing same issue. I am using Docker Desktop 3.2.2 with Docker Engine 20.10.5.
As soon as I spin any container my mac fans spin like anything and CPU ramps up to 153%. What is this com.docker.hyperkit??

I am using 2017-mac with MacOs BigSur v11.2.3. As this process drawing most of the CPU usage, my intelliJ is crashing and running like a slug. Does it make any difference if I uninstall docker desktop and install it via brew. But, I don’t see any point in going that path though.