[Cluster-devel] Re: Groupd uevent clean up

[Steven Whitehouse <swhiteho@redhat.com>]

Hi,

On Tue, 2008-11-25 at 11:08 -0600, David Teigland wrote:
> On Tue, Nov 25, 2008 at 04:44:32PM +0000, Steven Whitehouse wrote:
> > > OK, thanks, I dislike the const sprinkling, and I don't care about string
> > Whats wrong with adding the consts? It seems to make sense to mark the
> > fsname as const since its just being used as a key for looking up the
> > mg.
> 
> The PITA factor outweighs it's advantages, at least in the code I have to
> maintain myself.
> 
> > My dev machine currently has a strange mix of F-9 and (old) rawhide
> > packages on it, so that installing the required bits of openais to build
> > isn't so easy. If you are able to run a quick test, that would be
> > helpful otherwise, I'll look for another place to build it for testing.
> 
> I'll test a new patch that's limited to the parsing change.
> 
> > I've just noticed dlm_controld has this same bug,
> 
> Fixed that one too.
> 

So I think this needs looking into some more... I've pushed patches to
the tree which send more useful information with the uevents so that we
can avoid reading the sysfs files in many cases, in future. Each uevent
from gfs[2]/lock_dlm now includes:

LOCKTABLE=<clustername:fsname>
LOCKPROTO=[lock_dlm|lock_nolock]

to avoid all the messy parsing of the initial event string. Also I've
added come further information to the two "change" events, so that we
now have:

FIRSTMOUNT=Done

when the first mounter has finished and:

JID=<journal_id>
RECOVERY=[Done|Failed]

when recovery has finished. Is there anything else that is useful I
wonder, while I'm adding new items here? I think I've covered the
important bits anyway.

So I need to revise the initial patch a bit in order to take into
account the new information,

Steve.