I’m running into an issue where on reboot instances sometimes fail to start because a mount isn’t available:
level=warning msg="Failed auto start instance attempt" attempt=3 err="Failed to start device \"data\": Failed accessing source path \"/home/adgh/containers/adgh\": stat /home/adgh/containers/adgh: no such file or directory" instance=adgh maxAttempts=3 project=default
For context /home/adgh/containers/adgh is a symlink to a data folder (also on device) but doesn’t seem to be immediately accessible. (not sure why it takes so long for on-device symlinks to become live, but w/e)
I worked around the issue with Podman by taking advantage of the Quadlet systemd integration by specifying RestartSec=5 and the default systemd retry number of 5 seems to be enough.
I see from the logs that Incus’ delay between retries is 5 seconds, but it only tries 3 times (per the maxAttempts=3 and the last log entry says attempt=3).
I cannot see in the docs how to specify a higher maxAttempts or a higher time between start up attempts. Based off of systemd, the mount should be live by 20 seconds, so if I can tell it to try till then, that would fix it :\)
Final context, I have these boot settings set:
boot.autorestart: "true"
boot.autostart: "true"
boot.autostart.priority: "1"
I’d happily take setting boot.autostart to use boot.autorestart’s “10 tries in 1 minute” per Instance options - Incus documentation
Thanks! I’ve been using Incus for quite a while and it is very reliable, knowing this bit will make it perfect (for me at least!)