[master] Prevent stupid hang and 100% CPU usage after it

David Lehman dlehman at redhat.com
Mon Sep 10 14:58:17 UTC 2012


On Mon, 2012-09-10 at 14:21 +0200, Vratislav Podzimek wrote:
> PATCH 1/2 replaces GLib.idle_add with GLib.timeout_add with 250ms interval
> which is enough short to display changes and enough long to not cause high
> CPU usage.
> 
> PATCH 2/2 adds one more option to tell 'lvm' command that we really want
> to do what we are telling it to do. However, I'm not sure if this all is
> right. We need to use --force --force --yes because we are removing PV
> before removing VG that uses it. But since this blocks kickstart
> installations using 'clearpart --all', I'd like to push this at least as
> a temporary solution.

It used to be that lvm/md/whatever were not active until/unless we
activated them. Now there is all kinds of stuff (dracut, systemd to name
just two constant/repeated offenders) that really really wants to make
sure that everything gets automatically activated during system startup.
This has been "fixed" several times in the past but upstream keeps
un-fixing it, seemingly each fedora release. I am on the verge of
writing code to deactivate all this crap ourselves so we can stop
worrying about what those before us do.

Dave



More information about the anaconda-patches mailing list