Quoth Monte Pickard on Thu, Apr 27, 2006 at 01:38:50PM -0700: > For the TX (Trusted Extensions) test suite, we need to restart the > 'system/name-service-cache' service when certain system files are > changed. This is done with the following command: > > /usr/sbin/svcadm restart system/name-service-cache > > Some times that service goes into 'maintenance', with the message: > > [ Apr 26 22:05:51 Restarting too quickly, changing state to maintenance ] ... > What is the expected result if that command is executed 'back to back'? > Is there 'protection' from a second call interfering with the first? > Does the service have to 'protect' itself?
I believe this is simply a matter of svc.startd's algorithm being too simplistic. It should exempt svcadm restart-induced restarts from the restarting-too-quickly logic. File a bug. > If this is considered a 'no-no', then what can the test cases do to > avoid causing the 'maintenance' mode here? Is there a 'state' or > something that can be checked before issuing the 'restart'? You can sleep 1 or 2 between the restarts. David
