docs: Avoid memory bloat in some misconfigured versions of Docker/containerd Hosts (e.g. on Debian 13)

[dyxyl]

As found in #3045 and #4940, some versions of Docker/containerd set the "nofiles" limit very high, which causes beam.smp to use > 2 GB of memory. In particular, this happens in the current Debian stable (13/Trixie).

I've checked that this reduces the memory size in the default docker install, but not the Apache or Traefik installs.

[netlify]

✅ Deploy Preview for teslamate ready!

Name	Link
🔨 Latest commit	9f19d9d
🔍 Latest deploy log	https://app.netlify.com/projects/teslamate/deploys/6902149e7feea900095396b1
😎 Deploy Preview	https://deploy-preview-4999--teslamate.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[JakobLichterfeld]

Thanks for your contribution.
Generally speaking: configuring the host or docker is not in the scope of TeslaMate. There are a ton of useful docker options, but they would blow up the docker guides.

[brianmay]

I think it might be worth while here considering how many people affected. And we already do document the docker-compose file.

[JakobLichterfeld]

I tend to agree since the beginning.
From a beginner's perspective, these values seem to appear out of nowhere without any explanation. This does not inspire confidence.

[JakobLichterfeld]

To explain this to newcomers, a comment would greatly enhance UX.

[JakobLichterfeld]

To explain this to newcomers, a comment would greatly enhance UX.

[JakobLichterfeld]

To explain this to newcomers, a comment would greatly enhance UX.

[dyxyl]

I agree about the cryptic nature of these settings.

Please consider the change I've just made to entrypoint.sh instead, so the docker-compose file changes are not needed.

[swiffer]

If the change to the entrypoint alone is good enough i would vote for that.

Lower overall maintenance for us and less complexity exposed to users.

Could you update the pr and remove the other changes?

[JakobLichterfeld]

If the change to the entrypoint alone is good enough i would vote for that.

I do see it differently. This is hiding a setting which is host specific and even worse make it hard coded and not changeable. In my instance the ulimits is much higher and this change will decrease performance.

So I do see several options:

• an env variable and default to 65536. Which will decrease performance for some users
• using docker setting, as it is most likely caused by misconfigured docker hosts (and add proper comment). This has the drawback of more maintenance for us and more complexity exposed to users.
• do not change anything, as it is not TeslaMate related

[swiffer]

is that really affecting performance?

• i read it as "maximum number of simultanously open ports" - and ERL_MAX_PORTS 65536 sounds a lot.

[JakobLichterfeld]

To add: I strongly vote against using a deprecated env variable:

"Previously the environment variable ERL_MAX_PORTS was used for setting the maximum number of simultaneously existing ports. This environment variable is deprecated, and scheduled for removal in OTP-R17, but can still be used."

source: https://erlang.org/documentation/doc-7.2/erts-7.2/doc/html/erl.html

[swiffer]

To add: I strongly vote against using a deprecated env variable:

"Previously the environment variable ERL_MAX_PORTS was used for setting the maximum number of simultaneously existing ports. This environment variable is deprecated, and scheduled for removal in OTP-R17, but can still be used."

source: https://erlang.org/documentation/doc-7.2/erts-7.2/doc/html/erl.html

fully agree, wasn't aware. in that case also voting for nothing or adding it to the docs.

[dyxyl]

Ah, I didn't realize the env var was deprecated. I've just pushed a change to set the ulimit directly. I don't see how this will decrease performance for anyone (assuming no one has thousands of cars connected to the same instance).

[dyxyl]

I've refined entrypoint.sh again with a configuration variable and also to avoid an error if the system's limit is smaller.