Service Management Facility (SMF) - Service Won't Go Online

Question

Hello All,
I am new to this forum, but getting rather frustrated and thought I'd see if any of you could point me in the right direction.
I have just loaded Solaris 10 on my local workstation and I'm trying to get it set up. I have a simple script that I did have set up in my Solaris 9 inittab file to respawn. This script essentially kept track of my filesystem size and sent me an email if any specified filesystem went over a defined limit. It worked great for me in my Sol9 environment.
I have been trying to create a service (manifest?) for this same script, and I am clearly doing something wrong because when I import/enable the xml file, the service runs (I can see the process consistantly running, exiting, restarting), but the service state is offline.
I confess, this is my first attempt at a service script...and I can't seem to find much on-line info about the values - I have essentially just copied bits and pieces from other service xml files.
Please take a look and let me know if you see anything obviously wrong.
# svccfg import fscheck.xml
# svcs -a |grep fscheck
offline* 12:53:25 svc:/site/fscheck:default
# svcs -xv fscheck
svc:/site/fscheck:default (?)
State: offline since Tue Jul 24 12:53:25 2007
Reason: Start method is running.
See: http://sun.com/msg/SMF-8000-C4
See: /var/svc/log/site-fscheck:default.log
Impact: This service is not running.
# ps -ef |grep fscheck
root 3184 818 0 11:12:01 pts/4 0:00 tail -f site-fscheck:default.log
root 3484 7 0 13:08:26 ? 0:00 /bin/ksh /lib/svc/method/fscheck
Output of log file.
[ Jul 24 12:56:25 Method "start" exited with status 0 ]
[ Jul 24 12:56:25 Stopping because all processes in service exited. ]
[ Jul 24 12:56:25 Executing stop method (null) ]
[ Jul 24 12:56:25 Executing start method ("/lib/svc/method/fscheck") ]
# svcs -l fscheck
fmri svc:/site/fscheck:default
enabled true
state offline
next_state online
state_time Tue Jul 24 13:08:26 2007
logfile /var/svc/log/site-fscheck:default.log
restarter svc:/system/svc/restarter:default
contract_id 214
dependency require_all/none svc:/milestone/multi-user-server (online)
# more fscheck.xml
<?xml version="1.0"?>
<!DOCTYPE service_bundle SYSTEM "/usr/share/lib/xml/dtd/service_bundle.dtd.1">
<service_bundle type='manifest' name='manifest:fscheck'>
<service
name='site/fscheck'
type='service'
version='1'>
<create_default_instance enabled='true' />
<single_instance />
<dependency
name='foo_opt'
type='service'
grouping='require_all'
restart_on='none'>
<service_fmri value='svc:/milestone/multi-user-server' />
</dependency>
<exec_method
type='method'
name='start'
exec='/lib/svc/method/fscheck'
timeout_seconds='0' >
</exec_method>
<exec_method
type='method'
name='stop'
exec=':true'
timeout_seconds='0' >
</exec_method>
<property_group name='startd' type='framework'>
<propval name='duration' type='astring' value='wait' />
</property_group>
<stability value='Unstable' />
</service>
</service_bundle>

Answer 1

It looks like your startup script is hanging somehow. Run a ps and grep the process of the startup script to see if it has any children. That may help you determine where it is getting stuck. If you can't find any children, edit that script and add checkmarks that write to a debug file. Things like:
echo "checkpoint 1: `date` " >> /tmp/fscheck
Or... now that I think about it.... did you simply put your script in place at:
/lib/svc/method/fscheck ??
If so, that's your problem. SVC expects the startup script to finish, and from there tracks the children.
Put fscheck somewhere else, and make /lib/svc/method/fscheck a script that calls your fscheck as a background process.