Hi,

On a vanilla 138 install it seems that if I simulate a service
restarting too often (by killing processes) then it enters
maintenance state as expected, but thereafter a 'svcadm clear ...'
does not succeed - it just returns to maintenance state as before.
If I svcadm disable then svcadm enable it clears ok, which is
pretty much what I understood 'clear' to mean.

I've tried ntp and intrd services - same behaviour.

(gav...@enogas:~ )-> uname -a
SunOS enogas 5.11 snv_138 i86pc i386 i86pc

(gav...@enogas:~ )-> svcs intrd
STATE          STIME    FMRI
online         13:03:21 svc:/system/intrd:default

(gav...@enogas:~ )-> pfexec pkill intrd
(gav...@enogas:~ )-> pfexec pkill intrd
(gav...@enogas:~ )-> pfexec pkill intrd
(gav...@enogas:~ )-> pfexec pkill intrd

(note: not scripted - I am stopping the kills!)

And, as expected:

(gav...@enogas:~ )-> svcs -xv intrd
svc:/system/intrd:default (interrupt balancer)
 State: maintenance since Thu May 06 13:10:49 2010
Reason: Restarting too quickly.
   See: http://sun.com/msg/SMF-8000-L5
   See: man -M /usr/share/man -s 1M intrd
   See: /var/svc/log/system-intrd:default.log
Impact: This service is not running.

The log shows:

[ May  6 13:10:49 Stopping because all processes in service exited. ]
[ May  6 13:10:49 Executing stop method (:kill). ]
[ May  6 13:10:49 Restarting too quickly, changing state to maintenance. ]

But now I attempt to clear:

(gav...@enogas:~ )-> svcadm clear intrd

(gav...@enogas:~ )-> svcs -xv intrd
svc:/system/intrd:default (interrupt balancer)
 State: maintenance since Thu May 06 13:13:30 2010
Reason: Restarting too quickly.
   See: http://sun.com/msg/SMF-8000-L5
   See: man -M /usr/share/man -s 1M intrd
   See: /var/svc/log/system-intrd:default.log
Impact: This service is not running.

Still in maintenance.  Note the time is updated.  Addition log file
entries are:

[ May  6 13:13:30 Leaving maintenance because clear requested. ]
[ May  6 13:13:30 Enabled. ]
[ May  6 13:13:30 Restarting too quickly, changing state to maintenance. ]

Now that's all on vanilla bits.  On a lab machine running some project
bits I see the same behaviour, but there I have some functionality that records
graph transitions.  We do see a transition maintenance -> uninitialized
followed by uninitialized -> offline for the clear request, but the
next transition is offline -> maintenance for restarting too quickly.
I'd guess that the old baggage of the restart count/frequency is not being
cleared on this path??

Gavin

_______________________________________________
smf-discuss mailing list
[email protected]

Reply via email to