I am just installing with helm. Any ideas what is causing this?

I did

helm update

that might be it?

How much bandwidth does an average browser based user workspace use?

how does coder handle resource allocation, does it automatically allocate more cpu/ram under more intensive workloads? I am running the default docker template on a server with 56 threads and using a single workspace, jetbrains reports 97% CPU usage while indexing my project with 2 java files

it depends on the docker provider

coder doesn't handle allocation

I see

see ``cpu_set``, ``cpu_shares``, ``gpu``, ``memory``

I'm alone on an AMD EPYC 7502P with 128G ram, and intellij also reports 98% CPU frequently. I have no idea where it gets these numbers from, but the node underneath doesn't really reflect this. Maybe its "load" - that would go up if there is IO-Wait 🤷

@kyle Hey, so I added the wildcardhost as you suggested, but code-server port forwarding is still going to localhost. Any suggestions?

hey, I think this would be better in #1006346052317753414

I'll report it there. Thanks :))

I think docker restricts the computing power by default

https://docs.docker.com/config/containers/resource_constraints/#cpu "By default, each container’s access to the host machine’s CPU cycles is unlimited."

@JanL could you try experimenting with cpu options and report if they do any good ?

My comment was more like "I do not know how jetbrains measures the CPU load". I know that I have no limitations. The options the TF provider gives wont help anyway. There is only "shares" and "set" - which translates to the scheduling weight (shares) - so what "fraction" of your compute power you get if the system is under load - and the set on which cpu the container processes are allowed to run (meaning that if you have 32 cores and you set there 1,2,3, then it can run not just only on 3 cores but specifically on core 1,2 and 3. This is very rarely what you want and helps with usecases that suffer from cpu-cache-issues or interrup routing)

I believe by default docker should not restrict resources until the system starts OOM killing processes or the like, so I agree with @JanL that it might be an issue with how Jetbrains calculates CPU usage. I have no problem if only the counter is incorrect, but I am not sure how to test if Jetbrains temporarily stops indexing or other important background processes if it determines that it is out of resources.

iirc "load" can go above 100%

yeah, load roughly "tasks waiting"

If you really want to know if compute/memory/io is a bottleneck, linux provides pressure-stall-information for recent kernels.

https://facebookmicrosites.github.io/psi/docs/overview has some more visual explaination

I always see #cpus as 100% load. So on a 32 core system, a load of 32 would be 100% because there are 32 tasks waiting for 32 cpus - and this can grow indefinitely - but it's inaccurate as not all cores are equal, IO pressure has a huge effect on these numbers, etc. So, it's a more rough estimate how your system is overall: load < #cpus - you're fine, load > #cpus - you might consider getting a better machine. But for these conclusions, look on the long-term load average, as spikes are normal, especially if you have varying IO patterns.

I saw that you can block downloads from vs code in the browser but does anyone know if a user would be able upload files? Can that also be blocked?

Is this possible on free version?

wait for this PR https://github.com/coder/coder/pull/4311

ok