CC'ing nwam-discuss for broader audience

Renee Danson Sommerfeld wrote:
> As I mentioned in my comments on this bug
> (http://defect.opensolaris.org/bz/show_bug.cgi?id=11082), I'm
> concerned that the enable-svcs and disable-svcs properties are
> likely to cause confusion and inconsistent behavior for people
> who try to use them.  We don't track the state of the service
> prior to activation of the location, so we don't know how to
> restore that state when the location is deactivated.
>
> Furthermore, users can get really similar functionality by
> creating ENMs that are dependent on the location.
>
> So we have a feature that's both somewhat redundant, and also
> difficult to implement in a way that will result in consistent,
> predictable behavior.  That's not a good combination.
>   
The portion that's missing from the ENM is the equivalent of 
"svcs-disable".  "svcs-enable" can be reworked by having an ENM for each 
service and the ENM having a conditional activation on the location 
being active.  However, the services that need to be disabled when a 
location is active is not straightforward.  Users will have to create 
script(s) with "svcadm disable" and create ENM(s) for it(them).  Not a 
total loss, so I am ok with removing the two "svcs-enable" and 
"svcs-disable" properties.

> I also think that in the future, as locations become more
> smf-connected, adding this functionality, and making it work
> well, will be much easier.
>   
By the way, ENMs also have this problem.  If a service is already active 
before an ENM containing that service is enabled, then that service is 
disabled when the ENM is disabled.

I am getting confused now and am wondering if we are talking about 
removing the "fmri" property from ENM also (along with the location 
properties).

Anurag

Reply via email to