[Spacewalk-list] Modification of spacewalk-service script

[Jan Pazdziora]

On Sat, Nov 12, 2011 at 09:52:07PM +0100, Pierre Casenove wrote:
> Ok no problem, no patch so. If you're against it, I won't insist!
> Just to answer your email : HTTPD is also quite independant from the
> spacewalk product and still, it is controlled. And if postgresql gets

It is not as independent as we'd like -- we use mod_wsgi/mod_perl, so
for example an upgrade of Spacewalk requires restart of httpd.

On the other hand, even upgrade of Spacewalk's database schema does not
require restart of the underlying database server.

> restarted with SP up and running, spacewalk needs also to be restarted
> (connection is lost and never gets restored)

This would be a bug, it should be reported as such, investigated and
fixed.

On Sat, Nov 12, 2011 at 04:54:48PM -0500, Miroslav Suchy wrote:
> ----- Original Message -----
> > I'm against this change. The database (PostgreSQL, Oracle XE) is
> > completely independent from the Spacewalk software product and the
> > spacewalk-service script should have no business fiddling with them.
> 
> Define "completely independent". It is not true. Spacewalk use it. As it use apache.
> And tomcat. And jabber.

That's true, Spacewalk uses the database. But the database server is
like filesystem or network for Spacewalk. No action of the Spacewalk
application (not even its upgrade) requires restart of the database.
Unlike apache, tomcat, and (presumably) jabberd -- we modify the
configuration of apache, tomcat, and jabberd for Spacewalk to work
correctly, and for that we need to restart those services.

> I - personally - would not spent my devel cycles on it, but Pierre is willing
> to do that and contribute it.
>
> We recommend to run Spacewalk on dedicated machine. And I suppose that most users
> obey this recommendation. So if there is DB instance, it is very likely dedicated
> to Spacewalk.
> But yes, it can be used for other purposes as well. For that reason, I wanted to
> have this option disabled by default. So current behavior will be preserved.

[...]

> As I said, more user already wanted it. And it is much easier and better for users 
> to code this wrapper once, and leave dozen admins to flip one variable as compared
> to leave dozens admins to write theirs wrapper.

The problem is -- it is likely to become maintenance burden in the
long run. Noone but few people will use it because the default will
be to have the option off, therefore we won't be testing it.

-- 
Jan Pazdziora
Principal Software Engineer, Satellite Engineering, Red Hat