[OpenIndiana-discuss] Bug or feature in SMF? "svcadm restart" vs "disable+enable"

[Jim Klimov]

On 2013-11-18 21:22, John D Groenveld wrote:
> In message <528A74A1.7010008 at cos.ru>, Jim Klimov writes:
>> Yes, of course. Being a service under its local zone's SMF control,
>> the database starts and stops correctly. And so does the appserver
>> in its zone. The problem is that when the appserver-zone depends on
>> the database-zone, and I do "svcadm restart database-zone" - this
>> zone does stop, and only afterwards does the appserver-zone begin
>> to stop (and the database-zone begins to start in the meanwhile).
>> This is different from "svcadm disable (-t) database-zone" which
>> stops the appserver-zone first and then begins the shutdown of
>> database-zone, and then restarts them both from cold state.
>
> Do you have a plan to handle the app server and the database
> server running on separate hardware?

I see where you're going with this, and it is a good subject in its
own (clusterware, scripts and other signalling "I'm gonna shut down,
all clients go sleep before you die"), but in this case the DB-APP
interaction was presented as an example. There's any number of good
reasons to want dependencies to stop and start in defined order in
any case of controlled shutdown or restart. Heck, that's why we do
define them - and that is the subject of this discussion :)

For the most part our deployments were presented to customers and
in-house users as single-box appliances specifically to avoid this
problem of ungraceful shutdowns, at least as much as we can avoid it.
When customers wanted to grow into redundancy or otherwise scaled
behavior, this was a separate task to tackle (i.e. with replicated
databases, with primary sources still local to the applications).

In particular, zones that were too critical for ability of others
to work (say, the "hardware-local" replica of LDAP server with
definitions of POSIX users who own processes in other zones via
ldap/client, or of administrative accounts in the managed enterprise
applications - without an LDAP connection you can't even login into
the app to tell it to go down) might be managed by a suite of scripts
that started and stopped zones in a particular order. In a way, this
was indeed a home-made SMF built back in the days before we grasped
the real thing :)

And so far, regarding this restart bug, it still performs better ;P

//Jim