I just realised I still haven't replied to this. Sorry for the delay. On Mon, 2011-03-07 at 17:54 +0100, Luk Claes wrote: > On 03/06/2011 07:37 PM, Ben Hutchings wrote: > > On Sun, 2011-03-06 at 18:41 +0100, Luk Claes wrote: > >> On 03/06/2011 06:11 PM, Ben Hutchings wrote: > > >> I'm willing to help out. What bugs or area do you want me to prioritise on? > > > > I think the most urgent would be reported regressions from lenny to > > squeeze. > > I had a look at the outstanding bugs. Unless there are kerberos or nfs4 > related regressions, there don't seem to be real regressions reported. > > The issues that are reported quite a lot are problems related with the > init scripts though. Do you mind if I try to come up with some init > script changes that make sure depended on services are tested for > availability and services don't hang unnecessary? So you've done some of this, but I don't think you've gone far enough yet. Per #623377, I think we need to split nfs-common.init into at least: 1. rpc.statd (start in S, stop never; doesn't require /usr) 2. rpc.idmapd and rpc.gssd (start in 2345, stop in 016; requires /usr) Although mrvn reckoned idmapd might be needed for mounting in /usr. Maybe we should move everything from /usr to /. > I have the feeling that changes there would be welcomed most by our > userbase as I can read quite some frustration in the bug reports. > > Cheers > > Luk > > PS: The reason so many bugs are reported related to initscripts very > probably are a result from the different behaviour of the init system > (running things in parallel) and hotplug system (background handling if > possible) and/or cleaning up of other init scripts which cause faster > execution of our init scripts. The whole idea of treating network configuration as a single service ($network) is broken. For now, we just have to say that mounting at boot requires that the network is configured as auto and not hotplug. Really, it ought to be possible to define that NFS mounts depend on specific networks being available (though unmounting when the network has already gone away is... painful). Ben. -- Ben Hutchings Computers are not intelligent. They only think they are.
Attachment:
signature.asc
Description: This is a digitally signed message part