[rescue] Solaris 11 SMF oddity

[Phil Stracchino]

I have a minor puzzle. My X4540 rebooted itself yesterday.

The first problem with this is that I haven't been able to figure out
why, and it didn't log anything. But it appears to have just possibly
been something done (without asking for permission) by
usr/lib/update-manager/update-refresh.sh.

The second problem is that after it came back up, Bacula wasn't running,
and with no evidence of anything wrong with it, wouldn't restart ...
until I re-imported its manifest into SMF.

It seems like SMF had the Bacula services in its running instance, but
hadn't actually committed it. So now I'm looking for a subcommand to
tell svccfg to save or commit the imported service definitions ... and I
can't find one.


Does anyone else who runs Solaris 11 know if I'm missing something here?

[Edward Mitchell]

Sent from my iBrain.

Post by Phil Stracchino
I have a minor puzzle. My X4540 rebooted itself yesterday.
The first problem with this is that I haven't been able to figure out
why, and it didn't log anything. But it appears to have just possibly
been something done (without asking for permission) by
usr/lib/update-manager/update-refresh.sh.
The second problem is that after it came back up, Bacula wasn't running,
and with no evidence of anything wrong with it, wouldn't restart ...
until I re-imported its manifest into SMF.
It seems like SMF had the Bacula services in its running instance, but
hadn't actually committed it. So now I'm looking for a subcommand to
tell svccfg to save or commit the imported service definitions ... and I
can't find one.
Does anyone else who runs Solaris 11 know if I'm missing something here?
_______________________________________________
rescue list - http://www.sunhelp.org/mailman/listinfo/rescue

Hmm. svccfg import should be all it takes.

What does svccfg show, if anything, for snapshots of the Bacula FMRI?

And no gripes about a busted boot archive on the reboot?

FWIW, I've had similar reboots on X3-2's due to silently intermittent power
supplies.
_______________________________________________
rescue list - http://www.sunhelp.org/mailman/listinfo/rescue

[Phil Stracchino]

Post by Edward Mitchell
Sent from my iBrain.

Post by Phil Stracchino
Does anyone else who runs Solaris 11 know if I'm missing something here?

Hmm. svccfg import should be all it takes.

That's what I thought.

Post by Edward Mitchell
What does svccfg show, if anything, for snapshots of the Bacula FMRI?

asgard:root:~:10 # svccfg -s bacula/sd:default
svc:/network/bacula/sd:default> listsnap
initial
previous
running
start

Post by Edward Mitchell
And no gripes about a busted boot archive on the reboot?

I didn't see anything logged, and didn't see the console because I
didn't actually know it rebooted until later when I started
investigating why Bacula wasn't running.

[Jerry Kemp]

Hello Phil,

I'm sure you have probably already checked this, but I am curious anyway,
specifically regarding your "bacula" service.

I'm sure that any Bacula daemons would create log files that would probably
provide some assistance and insight, but I am specifically wondering if you have
checked the "bacula service" log file.

i.e., when you are logged in as root, and do something like/similar to this:

# svcs -l bacula

typically, the 4th line from the bottom list the location for the log file for
the bacula service.

More often than not, if the daemon log file itself won't provide specific
insight as to why it won't stay running, the service fault management log file
will typically provide insight as to why the daemon won't stay running after the
fault management services try (and fail) xxxx number of times.

Good luck,

Jerry

Post by Phil Stracchino
I have a minor puzzle. My X4540 rebooted itself yesterday.
The first problem with this is that I haven't been able to figure out
why, and it didn't log anything. But it appears to have just possibly
been something done (without asking for permission) by
usr/lib/update-manager/update-refresh.sh.
The second problem is that after it came back up, Bacula wasn't running,
and with no evidence of anything wrong with it, wouldn't restart ...
until I re-imported its manifest into SMF.
It seems like SMF had the Bacula services in its running instance, but
hadn't actually committed it. So now I'm looking for a subcommand to
tell svccfg to save or commit the imported service definitions ... and I
can't find one.
Does anyone else who runs Solaris 11 know if I'm missing something here?

_______________________________________________
rescue list - http://www.sunhelp.org/mailman/listinfo/rescue

[Phil Stracchino]

Post by Jerry Kemp
Hello Phil,
I'm sure you have probably already checked this, but I am curious anyway,
specifically regarding your "bacula" service.
I'm sure that any Bacula daemons would create log files that would probably
provide some assistance and insight, but I am specifically wondering if you have
checked the "bacula service" log file.

Unfortunately there was nothing enlightening there, and the services
would start perfectly well manually. The problem eventually appeared to
be that smf had somehow forgotten the service definitions and was trying
to start them incorrectly.

[Carl R. Friend]

Post by Phil Stracchino
Unfortunately there was nothing enlightening there, and the services
would start perfectly well manually. The problem eventually appeared to
be that smf had somehow forgotten the service definitions and was trying
to start them incorrectly.

Something here smells like an environment problem. When things
start up from SMF can all the needed libraries be found and read from?
This is a very common problem with self-compiled services that then
try to get integrated into SMF.

Cheers!

[Phil Stracchino]

Post by Carl R. Friend

Post by Phil Stracchino
Unfortunately there was nothing enlightening there, and the services
would start perfectly well manually. The problem eventually appeared to
be that smf had somehow forgotten the service definitions and was trying
to start them incorrectly.

Something here smells like an environment problem. When things
start up from SMF can all the needed libraries be found and read from?
This is a very common problem with self-compiled services that then
try to get integrated into SMF.

I'm not convinced that is the case here. I couldn't figure out why the
services were not starting via SMF although I could start them manually
with no errors whatsovever, *until* I restarted manifest-import, and
then suddenly the services not only worked again but auto-started.

It really seemed as though after reboot, SMF was trying to start the
Bacula services using the wrong service definition somehow.