[05:58] <twb> I told libvirtd that its sockets should be grouped by a group that comes from LDAP
[05:59] <twb> Unfortunately, libvirtd fails to start now because "runlevel 2" triggers before DHCP has finished coming up, so LDAP isn't available.
[05:59] <twb> What's the *right* way to tell libvirtd to wait until DHCP is done?
[06:03] <twb> This is in ubuntu 10.04, if that matters.
[06:07]  * twb guesses net-device-up IFACE!=lo is the filthy hack that everyone uses
[06:12] <twb> Nope, that doesn't work.  It comes up on eth0's state being up, rather than DHCP finishing -- I think.
[06:45] <SpamapS> twb: In Oneiric we'll be changing runlevel 2 to go after ifup -a has returned
[06:45] <twb> Good, I guess
[06:45] <SpamapS> twb: so I think if you change /etc/init/rc-sysinit.conf to 'start on filesystem and started networking' it should work.. or maybe 'stopped networking'
[06:46] <twb> obviously dhclient will background itself if it takes too long, but I'm prepared to live with 'stuff breaks" in that case
[06:46] <twb> SpamapS: I've spent an hour or two on this, so I'm giving up and just adding a dhclient exit hook that chgrps
[06:46] <SpamapS> oh you need libvirt to go after dynamic interfaces..
[06:46] <SpamapS> start on net-device-up IFACE=ethX
[06:46] <twb> I tried net-device-up IFACE!=lo and it didn't work
[06:47] <SpamapS> the only reason that shouldn't work is that you have multiple interfaces
[06:47] <twb> It still tried to bring it up immediately before dhclient
[06:47] <SpamapS> do you have a physical, static interface maybe?
[06:47] <twb> Ah, you're right, I am still bringing up the second interface
[06:48] <twb> But then why isn't it trying to bring libvirt-bin up AGAIN when the dhcp interface comes up?
[06:48] <twb> http://paste.debian.net/115826/ <-- interfaces, FYI
[06:48] <SpamapS> i want to add another event which is "all of the automatic interfaces are up" .. meaning all static, and all network-manager interfaces
[06:48] <twb> FWIW this is a server, so no NM
[06:48] <SpamapS> probably because it was in a goal of "start" but just thrashing, respawning
[06:49] <twb> I will try specifying a specific iface
[06:49] <SpamapS> even w/o nm you have dhclient backgrounding itself
[06:49] <twb> SpamapS: nod
[06:49] <twb> 15:46 <twb> obviously dhclient will background itself if it takes too long, but I'm prepared to live with 'stuff breaks" in that case
[06:49] <SpamapS> The question of "what are all of the automatic interfaces that we know about?" can be answered very easily.. once thats known.. we can trigger something to emit an event whenever all of those are up.
[06:51] <twb> I even tried "start on remote-filesystems", but that didn't work
[06:51] <twb> I assumed it would because mountall would send that after it finished mounting NFS
[06:51] <SpamapS> it most certainly does
[06:52] <SpamapS> IIRC it sends it when there are no remote filesystems.
[06:52] <twb> I do have NFS filesystems
[06:52] <SpamapS> And sometimes NFS isn't a "remote" filesystem.
[06:52] <twb> as a client, I mean, not a server
[06:53] <twb> omega:/home    /home    nfs intr,bg
[06:53] <twb> omega's a remote host, "omega" won't even resolve until DHCP is done
[06:54] <twb> OK, net-device-up IFACE=br-unmanaged seemed to work
[06:55] <twb> Thanks
[06:56] <twb> And sorry for losing my temper
[06:56] <SpamapS> Meh.. it does that. We need to put some fences around it.
[06:57] <SpamapS> also to be fair, your temper did not bleed through into IRC.. which is amazing.. usually an oz. of real temper becomes several metric tons of flames in the brain->keyboard->internet conversion process
[06:58] <twb> SpamapS: you aren't seeing the worst of it
[06:59] <SpamapS> twb: well I'm nodding off quite a bit.. need to either get caffeine or a pillow. :)
[06:59] <twb> No worries
[14:42] <dcorbin_wk> I want to write an upstart script for my service.  It should only start after certain old-style (init.d) scripts have run.  Do have to manually modify them to trigger upstart events?
[17:01] <SpamapS> dcorbin_wk: there are two ways to handle that..
[17:01] <SpamapS> dcorbin_wk: the best way would be to port those init.d scripts to upstart as well
[17:02] <SpamapS> dcorbin_wk: but if thats not possible, you can definitely add calls to 'initctl emit' to the init.d scripts that will help signal to your upstart job that its time to run
[17:29] <JanC> some services also provide "hooks" of course
[17:47] <Md> how can I debug why an ubuntu system with upstart hangs at boot? only plymouth and mountall are running
[17:51] <mbiebl> Md: you can use the --debug boot option
[17:51] <mbiebl> but I guess your problem is mountall related
[17:51] <Md> I did, but it did not help
[17:52] <Md> I have an open console and initctl list reports that plymouth and mountall are the only jobs running
[17:52] <mbiebl> do you have a special setup, like lvm, cryptsetup or something like this?
[17:53] <Md> no. actually this is the server of a customer
[17:53] <Md> he fucked up something while upgrading and asked us to install the last backup. which now does not work
[17:54] <Md> so for all I know the server could have been broken earlier, but I do not know where else to look
[17:54] <Md> mountall is just sitting there
[17:58] <mbiebl> I've seen something like that
[17:58] <mbiebl> a partition was defined in fstab
[17:59] <mbiebl> but the actual device was no longer attached to the server
[17:59] <Md> indeed that was it
[17:59] <Md> the swap partition now has a new uuid
[17:59] <mbiebl> and mountall was just sitting there waititing for the device to turn up
[17:59] <Md> this is fucking unbelievably stupid
[17:59] <mbiebl> yeah, I lost a ec2 instance that way
[18:00] <mbiebl> I cursed a whole day
[18:01] <mbiebl> I hope you have (physical) access to fixup the fstab
[18:03] <mbiebl> Md: that was some time ago, iirc a karmic installation. not sure if mountall is more clever nowadays and has timeouts 
[18:04] <dcorbin_wk> SpamapS: thanks.  I'm the init.d scripts are part of common ubuntu packages.  Minimal changes are preferred.
[18:05] <SpamapS> dcorbin_wk: on the contrary, we'd love to have things converted to upstart jobs. :)
[18:05] <dcorbin_wk> SpamapS: well, of course.  but this is "old ubuntu", well off of 'head'