We recently encountered this problem on one of our machines and decided to enable watchdog-timeout on the snap service just so that the service would be killed instead of triggering the system-level watchdog timer and there by avoiding device reboot.
The problem now is the service stays alive only for the duration of watchdog-timeout and gets killed at the end of timeout which was unexpected.
The question is do we have to implement something in our app to call sd_notify regularly as mentioned here or is this something the daemon-notify interface handles when we connect to it ?
Journalctl logs
Nov 15 09:17:18 systemd[1]: Started Service for snap application my-snap.myser
Nov 15 09:18:19 systemd[1]: snap.my-snap.myser.service: Watchdog timeout (limit 1min)!
Nov 15 09:18:19 systemd[1]: snap.my-snap.myser.service: Main process exited, code=killed, status=6/ABRT
Nov 15 09:18:19 systemd[1]: snap.my-snap.myser.service: Unit entered failed state.
Nov 15 09:18:19 systemd[1]: snap.my-snap.myser.service: Failed with result 'signal'.
Here is the relevant portion of the snapcraft.yaml
restart-condition: never
passthrough:
watchdog-timeout: 10m
restart-delay: 5s
Other info
$ systemd --version
systemd 229
+PAM +AUDIT +SELINUX +IMA +APPARMOR +SMACK +SYSVINIT +UTMP +LIBCRYPTSETUP +GCRYPT +GNUTLS +ACL +XZ -LZ4 +SECCOMP +BLKID +ELFUTILS +KMOD -IDN
$ snap version
snap 2.35.5
snapd 2.35.5
series 16
kernel 4.4.0-122-generic