Eventstore: Idle EventStore using 15-25% of CPU

[RichoDemus]

Hi, I downloaded the docker image eventstore/eventstore:release-3.9.3 and just started it up on my local machine (osx Sierra running Docker 1.13.1)

I can access the admin gui and I've tried some of the curl examples and Eventstore seems to be working, but as the title implies it's consuming an enormous amount of CPU for being idle. I don't know much about .NET or mono but I'd love to help troubleshooting this

EDIT: I tried downloading the package with mono from the website instead of running it as a docker container, it's a bit better but still consuming around 10% CPU

[shouze]

same thing for me

[cerlestes]

Starting eventstored after a fresh install, before any interaction with it, will make it use 5-10% CPU constantly on my Ubuntu 16.04 machine, too... There's nothing to do, why would anything run? This includes statistics - there are no events, why create statistics and waste CPU time?

[gregoryyoung]

there are things still running such as timing mechanisms, statistics, you
can see messages still being passed. It would be interesting to see a flame
graph

On Tue, Mar 21, 2017 at 9:53 AM, cerlestes [e-mail] wrote:

Starting eventstored after a fresh install, before any interaction with
it, will make it use 5-10% CPU constantly on my Ubuntu 16.04 machine,
too... There's nothing to do, why would anything run?

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
https://github.com/EventStore/EventStore/issues/1212#issuecomment-288028238,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAXRWvofLNiMCCFOBHRyryAbubcpHUe-ks5rn54vgaJpZM4MAraV
.

--
Studying for the Turing test

[gregoryyoung]

Try running the version coming out. I spent some time profiling around idle use. It should be dropped by 80-90%

[shouze]

@gregoryyoung related to #1282?

[pgermishuys]

@shouze yes

[hayley-jean]

This issue has been fixed by #1282

[gregoryyoung]

Can you guys verify that this PR resolves your issue?

[shouze]

@gregoryyoung yes I've built a .deb package & a docker image will make some tests tomorrow.

[gregoryyoung]

👍

[shouze]

Ok I've checked and this lead to a 3-4x improvement 🎉 Perfect timing to stay cool with my laptop during this summer 😉

[gregoryyoung]

Glad to hear it!

On Tue, Jun 6, 2017 at 11:36 AM, Sébastien HOUZÉ [e-mail]
wrote:

Ok I've checked and this lead to a 3-4x improvement 🎉 Perfect timing to
stay cool with my laptop during this summer 😉

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/EventStore/EventStore/issues/1212#issuecomment-306447601,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAXRWrmkW_p2FtvsMHK-47QqVphsoSG1ks5sBSu7gaJpZM4MAraV
.

--
Studying for the Turing test

[sledorze]

I start the latest docker image (4.1.1.0 - in memory) in Docker (Docker 18.05.0-ce-mac67) and my com.docker.hyperkit process is at a solid 28% CPU.
Any regression here? (Summer is Hot)

[sledorze]

@gregoryyoung should I create another issue ?

[bruth]

I am observing this today with the release-4.1.1-hotfix1 Docker image (pushed a month ago).

[sledorze]

@bruth I'm interested if you have a fix

[bruth]

I can't say I do. I have not dug into the code nor am I familiar with .NET, Mono, or C# in general.

[shaan1337]

@bruth hello, thanks for reporting this.

Can you please try the following:

1. Attach to the docker container
2. top -p <eventstored pid>
3. Press Shift-H - this should display the different threads
4. Can you please identify the thread that's consuming a lot of CPU? You should be able to grab the name from the Command column.

Thanks!

[bruth]

After starting up and nothing do anything with it for a few minutes, I see it idling around here:

• Index Committer ~ 4.5-5%
• Timer Scheduler ~ 2.5-3%
• Storage Chaser ~ 1-1.3%
• Timer ~ 0.5-1%

These seem actually reasonable on the surface?

The Docker hyperkit CPU util is misleading since I am seeing 20-30% and no other containers are running. So obviously that should not be the baseline since there are other factors at play.

[sledorze]

same as @bruth here..

[shaan1337]

@bruth @sledorze thanks for the feedback, I'll give this a try asap