Running out of space: Share same _work directory for all runners?

[Akronae]

Hi,

Thank you for your tool, it's really handy helping us manage ~30 runners.
One big issue I'm facing is that my 85GB SSD machines are quickly running out of storage.

This is mainly caused by the infamous node_modules folder being replicated for all runners (up to 8 on each machine).
It would be really helpful if those _work directories could be shared among runners.
In essence it could not only save space but also time I guess.

Do you think it could be doable?
I'd be happy to make a PR.

[vbem]

Hi Akronae,

I am very pleased to receive your feedback on usage.

Actually, the _work directory is the home directory of many workflow contexts, such as ${{ runner.temp }}, ${{ runner.workspace }}, ${{ github.workspace }}, etc. Sharing it among multi runners on the same host may cause conflicts of different workflow runs.

However, if the node_modules folder is the largest duplication occupant, maybe you can simply create a shared folder on the Linux host (e.g. /home/SharedFolder), and set the NODE_PATH: /home/SharedFolder/.node_modules environment variable in all related workflow files. Please note that as all runners created by this application were already in a shared Linux group runners, the /home/SharedFolder should also be owned by this group.

The disk usage may also increase because logs, workspace caches, runners updates, etc. You can also re-create the runners in crontab daily or weekly as a more complete housekeeping.

[Akronae]

Ok, thanks for your help, I chose the recreate all runners periodically as this will also fix the issue where half the runners disconnect and go offline after a month or so.

This seems to work fine:

#!/bin/bash
for i in {0..7}
do
  bash multi-runners/mr.bash del --user runner-$i
  bash multi-runners/mr.bash add --org <my org>
done

Cheers 😄

[vbem]

Awesome. Besides, starting from v1.2.0, del --org <your-org-name> --count 8 and add --org <your-org-name> --count 8 were also supported.