Hi,
On some of our production servers with a lot of JBOD disks, smartd
takes so long to start up that systemd's default 90s timeout kicks in
and smartd gets killed. Example log extract:
Mar 07 14:15:10 ms-be2070 systemd[1]: Starting Self Monitoring and
Reporting Technology (SMART) Daemon...
[smartd starts monitoring disks]
Mar 07 14:16:40 ms-be2070 systemd[1]: smartmontools.service: start
operation timed out. Terminating.
Mar 07 14:16:40 ms-be2070 systemd[1]: smartmontools.service: Failed with
result 'timeout'.
Mar 07 14:16:40 ms-be2070 systemd[1]: Failed to start Self Monitoring
and Reporting Technology (SMART) Daemon.
I fixed this by adding a systemd override:
[Service]
TimeoutSec=300
And now all is good:
Mar 07 14:31:53 ms-be2070 systemd[1]: Starting Self Monitoring and
Reporting Technology (SMART) Daemon...
Mar 07 14:34:06 ms-be2070 systemd[1]: Started Self Monitoring and
Reporting Technology (SMART) Daemon.
Note the ~2 minute start time.
It seems to me that the default startup timeout ought to be long
enough to work "out of the box", so could the unit file we ship be
adjusted to set a sensible TimeoutSec value, please? 300s seems likely
to be enough.
Thanks,
Matthew