On Thu, 2008-09-18 at 12:06 +1000, Max Matveev wrote:
> On Thu, 18 Sep 2008 11:41:18 +1000, Nathan Scott wrote:
>
> nscott> On Thu, 2008-09-18 at 11:31 +1000, Nathan Scott wrote:
>
> >> Does anyone know the reason for not starting pmcd after an rpm
> >> install? I've had a request for that, to reduce steps required
> >> to do a pcp install/upgrade on large numbers of machines ... is
> >> there any good reason not to do this?
> Part of it could be inertia - we've not done it before, why start now.
>
> But there is also the problem of installing pcp into chroot
> environment - you really don't want to be starting it there and it's
> most likely isn't going to run.
>
> If pcp were ever to be split into pcp proper, pcp-devel, pcp-eoe etc
> then the issue of installing the whole caboodle into chroot for builds
> is going to go away.
It might be time to do that. Be nice to have that split done
in the pre-3.0 timeframe, too. I'll think about taking that
on, maybe I can slot it in.
> nscott> Also, I just noticed a long-time PITA I've had, but never got
> nscott> round to looking into - the rpm postop explicitly chkconfig's
> nscott> pmie and pmproxy off. This means on an upgrade, previously
> nscott> enabled pmie and pmproxy state is blown away.
>
> nscott> Thats clearly a bug - I'll remove those two lines shortly in
> nscott> my branch.
> And replace it with what?
Those two lines (chkconfig_off calls in pcp.spec.in) don't need any
replacement - the chkconfig --add calls earlier, remain, so we just
don't touch pmie/pmproxy state at all (if it was on, it stays on; if
first install, they're off by default).
cheers.
--
Nathan
|