Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

UDEV and predictable resource ids #603

Open
febus982 opened this issue May 15, 2023 · 6 comments
Open

UDEV and predictable resource ids #603

febus982 opened this issue May 15, 2023 · 6 comments

Comments

@febus982
Copy link

Is there a way to have a predictable naming strategy for the the UDEV_DEVNODE and UDEV_DEVPATH environment variables?

Using the hash to differentiate the different devices works as long as it's possible to list all the environment variables and parse all of their prefix.

When you want to run containers with existing software (e.g. ser2net or zigbee2mqtt) it's difficult to specify the correct environment variable in the manifest because of the final name being unknown.

I am using a single ser2net pod for a single USB device configuration and I'd need to pass the exact variable name (or the variable content) to the podspec.

Right now I'm using this approach in a container containing the ser2net but it's kind of an hacky one.

CMD export ADAPTER=$(env | awk -F= '/^UDEV_DEVNODE/ {print $2;exit;}') \
 && echo "\
connection: &con01\n\
  accepter: tcp,20108\n\
  connector: serialdev,$ADAPTER,115200n81,nobreak,local\n\
  options:\n\
    kickolduser: true\
" > /ser2netconfig/config.yaml \
 && echo -n "Starting " && ser2net -v && ser2net -d -c /ser2netconfig/config.yaml

Also the same device changes suffix if attached to different nodes, so it's impossible to schedule pods using a separate manifest simply by requesting akri resources.

Is there a way to have predictable resource ids instead of a hash? Or at least an hash that depends only on the physical device information so it remains consistent if attached to different nodes/usb_ports

@diconico07
Copy link
Contributor

In order to have a predictable environment variable name you would need a predictable instance id.

However this would be difficult for udev as there are no unique identifier (other than node+device path), using something like vendor id/device id for a usb device wouldn't work if you have multiple of them.

For the "requesting akri resources" usecase, #565 will surely help you when out.

@febus982
Copy link
Author

That PR would be great! Can't wait to see it merged.

I don't want to add further requirements but would it make sense to use the suffix in the env variable only when more than one device is identified by the instance?

@jbpaux
Copy link
Contributor

jbpaux commented May 17, 2023

For your first issue you can try to write your /ser2netconfig/config.yaml file using in your podspec an init container responsible to get the variable and build the file

@kate-goldenring
Copy link
Contributor

That PR would be great! Can't wait to see it merged.

I don't want to add further requirements but would it make sense to use the suffix in the env variable only when more than one device is identified by the instance?

@febus982 this might be hard, but one thing we could do is always also set UDEV_DEVNODE and UDEV_DEVPATH for each device. It will be overwritten by each device and randomly contain one of the device's info, but for one device it would be accurate. A little hacky but maybe appropriate?

@kate-goldenring kate-goldenring moved this to Investigating in Akri Roadmap Oct 3, 2023
@github-actions
Copy link
Contributor

Issue has been automatically marked as stale due to inactivity for 90 days. Update the issue to remove label, otherwise it will be automatically closed.

@github-actions github-actions bot added the stale label Oct 23, 2023
@github-actions github-actions bot closed this as not planned Won't fix, can't repro, duplicate, stale Jan 21, 2024
@github-project-automation github-project-automation bot moved this from Investigating to Done in Akri Roadmap Jan 21, 2024
@github-project-automation github-project-automation bot moved this from Done to Triage needed in Akri Roadmap Feb 16, 2024
@yujinkim-msft yujinkim-msft moved this from Triage needed to Investigating in Akri Roadmap Mar 5, 2024
Copy link
Contributor

Issue has been automatically marked as stale due to inactivity for 90 days. Update the issue to remove label, otherwise it will be automatically closed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
Status: Investigating
Development

No branches or pull requests

4 participants