It is common to install a version from a branch or a fork. Change the upgrade to be a URL instead of a version number for simplicity.
Fix a bug by which the lxc-helpers and the runner are upgraded but not the script itself.
Increase the service unit timeout when stopping the runner: it may take up to three hours.
Adapt the tests accordingly.
Fixes https://code.forgejo.org/forgejo/runner/issues/510
Reviewed-on: https://code.forgejo.org/forgejo/runner/pulls/520
Reviewed-by: Michael Kriese <michael.kriese@gmx.de>
Co-authored-by: Earl Warren <contact@earl-warren.org>
Co-committed-by: Earl Warren <contact@earl-warren.org>
Install the version of the runner that the upgraded script needs.
Reviewed-on: https://code.forgejo.org/forgejo/runner/pulls/479
Co-authored-by: Earl Warren <contact@earl-warren.org>
Co-committed-by: Earl Warren <contact@earl-warren.org>
Once 6.2.2 is published, the example will need to upgrade to this version and not the previous one.
Reviewed-on: https://code.forgejo.org/forgejo/runner/pulls/478
Co-authored-by: Earl Warren <contact@earl-warren.org>
Co-committed-by: Earl Warren <contact@earl-warren.org>
for the purpose of indirectly upgrading lxc-helpers if a newer version is needed. Otherwise it has to be done manually and can easily be overlooked.
Reviewed-on: https://code.forgejo.org/forgejo/runner/pulls/475
Reviewed-by: Michael Kriese <michael.kriese@gmx.de>
Co-authored-by: Earl Warren <contact@earl-warren.org>
Co-committed-by: Earl Warren <contact@earl-warren.org>
re-creating the LXC container from scratch is expensive. When
rebooting or when multiple units start at the same time, it may cause
an execessive load.
Use a global lock to guard this operation so they happen in sequence
and not in parallel. They typically take around one minute to complete
which means that in the event of a reboot, it will take around 1
minutes * the number of runners for the unlucky one to start.
During this interval workflows will have to wait.
