IRC chat logs for #ltsp on irc.freenode.net (webchat)


Channel log from 28 November 2010   (all times are UTC)

00:08vagrantc_ has quit IRC
00:20vagrantc has joined #ltsp
00:33Damianos has joined #ltsp
00:33Damianos has left #ltsp
00:39sweetpi has joined #ltsp
01:01Kicer86 has joined #ltsp
01:18Kicer86 has quit IRC
01:24Kicer86 has joined #ltsp
01:29uday has joined #ltsp
01:36sweetpi has quit IRC
02:11gorkhaan has quit IRC
02:12gorkhaan has joined #ltsp
02:52
<alkisg>
stgraber: about the poweroff -fp and reboot -fp upstart scripts: aren't those just to prevent nbd-client from closing the nbd device?
02:52
/etc/nbd-client has an option to prevent this in a cross-distro way, without abruptly halting the computer and preventing other services from correctly closing down...
02:53
If KILLALL="false" and no NBD_DEVICE[0]=xxx devices are set in /etc/nbd-client, then everything works fine out of the box.
02:54
If we don't want to modify a conffile, we can just use >= /dev/nbd1 instead of nbd0, which is mentioned in that config file
02:54
(use nbd1 in ltsp, I mean)
03:00
We could probably tell Wouter, the nbd maintainer, to not include a NBD_DEVICE[0]=xxx entry in the default configuration file, that would be the best solution afaik
03:01
OK, filing two bugs for this, one in nbd and one in ltsp...
03:01
<vagrantc>
would even be better to not ship a default configuration file at all for nbd
03:01
seeing as there's no reasonable default
03:03
<alkisg>
vagrantc: would you like to file a bug in debian nbd, instead of me filing a bug in launchpad about it? It would be more direct...
03:04
<vagrantc>
alkisg: what's the exact situation?
03:04
alkisg: i didn't catch the entire conversation
03:05
<alkisg>
vagrantc: just what I wrote above, we didn't have a conversiation, but let me rewrite it better (gimme a few minutes)
03:09
<vagrantc>
in fact, it would be nice to get a /etc/nbd-client.d directory or some such
03:09
<alkisg>
dpkg-reconfigure nbd-client states:
03:09
"When the nbd-client init script is called to stop the nbd-client service, there are two things that can be done: either it can disconnect all nbd-client devices (which are assumed not to be in use), or it can disconnect only those nbd-client devices that it knows about in its config file."
03:09
The default in /etc/nbd-client is
03:09
KILLALL="false"
03:09
which means that the initscript should ignore any unknown nbd-client devices.
03:09
Unfortunately, in the same example file,
03:09
NBD_DEVICE[0]=/dev/nbd0
03:09
is used, so /dev/nbd0 is always killed even if the admin didn't manually specify any nbd connections.
03:09
That breaks shutdown in LTSP clients, it'd be better if no NBD_DEVICE was declared by default in the configuration file, or even if there was no configuration file at all shipped by default.
03:09
vagrantc: something like this ^^^ . stgraber tried to work around that problem with upstart scripts, but that breaks the shutdown process.
03:10
(so poweroff -fp should no longer be needed anywhere)
03:12dlezcano has joined #ltsp
03:15komunista has joined #ltsp
03:16
<vagrantc>
the sendsigs.omit stuff doesn't handle it? works on debian.
03:17
<alkisg>
vagrantc: which one? The one from nbd-client?
03:18
That would work only if nbd-client was started as a service
03:18
But we start it from the initramfs, so it doesn't put the devices in sendsigs.omit
03:19
I guess if we put "root=/dev/nbd0" in pxelinux.cfg and call "invoke-rc.d nbd-client start" as part of the ltsp client boot process, then it would work
03:20
vagrantc: ah, you mean the ones from ltsp-init-common? There's no need to duplicate the code afaik, that should be fixed in the nbd-client package
03:20
(and in the future, we can use the nbd-client initramfs hook for nbd root mounting, it works fine, there's no need to duplicate this code in ltsp_nbd either)
03:20vagrantc has quit IRC
03:21vagrantc has joined #ltsp
03:21
<vagrantc>
alkisg: gah. irssi froze on me...
03:21
<alkisg>
Oops
03:21
(11:17:54 AM) alkisg: vagrantc: which one? The one from nbd-client?
03:21
(11:18:11 AM) alkisg: That would work only if nbd-client was started as a service
03:21
(11:18:29 AM) alkisg: But we start it from the initramfs, so it doesn't put the devices in sendsigs.omit
03:21
(11:19:09 AM) alkisg: I guess if we put "root=/dev/nbd0" in pxelinux.cfg and call "invoke-rc.d nbd-client start" as part of the ltsp client boot process, then it would work
03:21
(11:20:07 AM) alkisg: vagrantc: ah, you mean the ones from ltsp-init-common? There's no need to duplicate the code afaik, that should be fixed in the nbd-client package
03:21
(11:20:32 AM) alkisg: (and in the future, we can use the nbd-client initramfs hook for nbd root mounting, it works fine, there's no need to duplicate this code in ltsp_nbd either)
03:22
<vagrantc>
anyways, on debian there are several things...
03:22
<alkisg>
Yet another way to fix it would be to add a _netdev_ line in /etc/fstab, the nbd-client initscript ignores those too
03:23
<vagrantc>
echo KILLALL=false >> $ROOT/etc/nbd-client ; echo OMITKILL=/dev/nbd0 >> $ROOT/etc/nbd-client
03:23
there's an ltsp-build-client plugin that adds those
03:23
basically works
03:24
<alkisg>
Hmm where? In the debian dir? I don't have those in my chroot...
03:24
Ah right, in Debian only
03:24
<vagrantc>
right
03:24
<alkisg>
$%^&* :)
03:24
<vagrantc>
well, while not default, sometimes i think debian has better support for NBD than ubuntu sometimes :)
03:24
<alkisg>
We need better sync between debian and ubuntu ltsp...
03:25
(as in, drop any other nbd stuff that ubuntu has and only keep the debian ones :D)
03:26
<vagrantc>
OMITKILL was added mid 2009
03:26
same with KILLALL=false
03:26
<alkisg>
vagrantc: unsetting NBD_DEVICE[0] should also work
03:26
instead of OMITKILL
03:28
But if Wouter removed NBD_DEVICE[0]=/dev/nbd0 from the default configuration file, then no LTSP workaround would be needed at all
03:29
<vagrantc>
so, the problem is that by default, it kills devices configured there, but there's no sane default configuration?
03:29uday has quit IRC
03:29
<alkisg>
Yes
03:32
<vagrantc>
that seems like a bug worth reporting
03:33dlezcano has quit IRC
03:34* alkisg reads at the nbd-client initscript code to find the best, less intrusive solution ...
03:35
<vagrantc>
alkisg: hm... at least in sid, nbd-client defaults to NBD_DEVICE[0]=
03:35
<alkisg>
vagrantc: I have squeeze here, I don't think the packages are different..?!
03:35
Ah no wait I have squeeze for the server
03:35
Was looking at the client chroot, which is an ubuntu one...
03:38* vagrantc is happy to hear that alkisg has some debian around
03:38
<alkisg>
vagrantc: yup, fixed in ubuntu natty too. So OK that would be a bug report worth filing if it wasn't already fixed :D
03:39
Now to report a bug about the upstart scripts in ubuntu-ltsp...
03:39
vagrantc: I've a thin client with 128 RAM in my home, I put squeeze / lxde on it, it's much lighter that anything ubuntu-based I found
03:40
I have it as a print server, scan server, nfs server, movie player for the kids etc, very handy and with a low power consumption
03:40
...and now I'm trying to make it an `nbd -c partition` server (for booting 1 single fat client without ltsp) as well ;)
03:41
The nbd-client package works fine for remote booting, we shouldn't need any nbd* related code in ltsp at all...
03:42
<vagrantc>
indeed it does.
03:42
at least, i've been using it
03:44
<alkisg>
Urm the squeeze package has "KILLALL=true" though :(
03:44
*sid
03:45
<vagrantc>
we could seed a debconf default
03:45
or manually tweak it
03:46
<alkisg>
I think the maintainer would accept code for nbd-client not to shut down network connections for /
03:47
So maybe the solution should be inside nbd-client, and not related to ltsp at all...
03:47
<vagrantc>
indeed, that would be cleaner
03:47
<alkisg>
Btw, /lib/init/rw/sendsigs.omit.d/nbd-client ==> shouldn't that be in /var/run?
03:48
<vagrantc>
alkisg: there are several different implementations
03:48
i tried to support as many as i knew about
03:48
<alkisg>
No I meant the code in the nbd-client initscript, not in ltsp
03:49
<vagrantc>
i know that works in debian
03:52
the /lib/init/rw thing was added a while ago to provide an early place to write stuff (since /var may be on a separate partition).
03:52
<alkisg>
Ah, so it's not obsolete... ok
03:53
So... I think that when /etc/init.d/nbd-client is invoked when the ltsp clients boot, it should check /proc/cmdline for the root fs, and put that in its omit list
03:54
Looking if that already works, or if it doesn't, why...
03:55
<vagrantc>
alkisg: were you able to get an aufs mount just using nbd-client's hooks?
03:55
<alkisg>
vagrantc: No i'm using a rw partition for nbd-server
03:55
I.e. a single fat client with no aufs
03:55
<vagrantc>
ah!
03:56* vagrantc did that recently with armhf tests
03:56
<vagrantc>
since qemu can't be used to emulate the new architecture
03:57komunista1 has joined #ltsp
03:57komunista has quit IRC
03:57komunista1 is now known as komunista
03:57yanu_ has joined #ltsp
03:57yanu has quit IRC
03:58chupadupa has joined #ltsp
04:00daduke__ has joined #ltsp
04:01chupacabra has quit IRC
04:01daduke_ has quit IRC
04:03BWMerlin_ has joined #ltsp
04:05
<alkisg>
I think nbd copy on write should also work, without aufs...
04:07
<vagrantc>
wouter said it was pretty slow
04:07
<alkisg>
Ah. Well it shouldn't be, so maybe some patches there would also be welcome... ;)
04:07BWMerlin has quit IRC
04:07ogra_ac_ has quit IRC
04:07gent_geen has quit IRC
04:07BWMerlin_ is now known as BWMerlin
04:07komunista1 has joined #ltsp
04:07komunista has quit IRC
04:07
<vagrantc>
good patches are generally welcomed :)
04:11gent_geen has joined #ltsp
04:15vagrantc has quit IRC
04:16MorningSon has quit IRC
04:18ogra_ac_ has joined #ltsp
04:44danalt has joined #ltsp
04:46danalt has quit IRC
04:47danalt has joined #ltsp
04:49danalt has quit IRC
04:50danalt has joined #ltsp
05:06
<alkisg>
Reported the nbd-client part of the problem at https://bugs.launchpad.net/ubuntu/+source/nbd/+bug/682328
05:24danalt has quit IRC
05:24danalt has joined #ltsp
05:28danalt has joined #ltsp
05:32alkisg has quit IRC
06:01BWMerlin has quit IRC
06:09vmlintu_ has quit IRC
06:33gorkhaan has quit IRC
06:33gorkhaan has joined #ltsp
06:44bobby_C has joined #ltsp
06:44dlezcano has joined #ltsp
06:50dlezcano has quit IRC
06:51gorkhaan has quit IRC
06:58gorkhaan has joined #ltsp
07:12Trixboxer has joined #ltsp
07:24Mobe_ has quit IRC
08:02johnny has left #ltsp
08:02johnny has joined #ltsp
08:13Kicer86 has quit IRC
08:22komunista1 has quit IRC
08:36komunista has joined #ltsp
08:57alkisg has joined #ltsp
09:19Trixboxer has quit IRC
09:59danalt has left #ltsp
10:01danalt has joined #ltsp
10:02danalt has quit IRC
10:23ogra_ac_ is now known as ogra_ac
10:34litlebuda has joined #ltsp
10:39Mobe has joined #ltsp
10:49komunista has quit IRC
10:52komunista has joined #ltsp
11:26pem__ has quit IRC
11:28vmlintu_ has joined #ltsp
11:37Kicer86 has joined #ltsp
11:46vmlintu_ has quit IRC
12:55mistik1 has joined #ltsp
12:58alkisg has quit IRC
12:59alkisg has joined #ltsp
13:14chupadupa is now known as chupacabra
13:37dlezcano has joined #ltsp
14:14litlebuda has quit IRC
14:21Selveste1 has quit IRC
14:21Selveste1_ has joined #ltsp
14:43shogunx has quit IRC
14:44shogunx has joined #ltsp
14:46chupacabra has quit IRC
14:54dobber_ has joined #ltsp
15:03Kicer86 has quit IRC
15:31vagrantc has joined #ltsp
15:31jhutchins_kc is now known as jhutchins
15:51alkisg has quit IRC
16:16bobby_C has quit IRC
16:44ltspbot has joined #ltsp
16:45ltspbot` has quit IRC
16:49johnny has left #ltsp
17:06komunista has quit IRC
17:12dobber_ has quit IRC
17:16Appiah has quit IRC
17:18Appiah has joined #ltsp
17:31ogra_ac_ has joined #ltsp
17:35ogra_ac has quit IRC
17:40shogunx has quit IRC
17:52johnny has joined #ltsp
18:04evil_root is now known as zz_evil_root
18:14shogunx has joined #ltsp
18:59mistik1_ has joined #ltsp
19:00mistik1 has quit IRC
19:00mistik1_ is now known as mistik1
19:18rjune has quit IRC
19:43chupacabra has joined #ltsp
19:50litlebuda has joined #ltsp
19:54Damianos has joined #ltsp
20:12johnny has left #ltsp
20:25ogra has quit IRC
20:25ogra has joined #ltsp
20:27vagrantc has quit IRC
21:02RiXtEr has joined #ltsp
21:29daya has joined #ltsp
21:31zz_evil_root is now known as evil_root
21:35evil_root is now known as zz_evil_root
22:05johnny has joined #ltsp
22:15litlebuda has quit IRC
22:18sweetpi has joined #ltsp
22:24alkisg has joined #ltsp
22:24alkisg has joined #ltsp
22:35sweetpi has quit IRC
22:35daya has quit IRC
22:47zz_evil_root is now known as evil_root
22:48evil_root is now known as zz_evil_root
22:54zz_evil_root is now known as evil_root
22:57blahal-rix has joined #ltsp
22:57RiXtEr has quit IRC
22:57blahal-rix is now known as RiXtEr
23:48vagrantc has joined #ltsp
23:55
<alkisg>
vagrantc: I believe that everything is working out of the box with the natty (and squeeze) nbd-client versions
23:55
So I think we can drop the poweroff -fp workarouds
23:55
...but we
23:56
...but we'll need to add "nbdroot=xxx" and "root=/dev/nbd0" in pxelinux.cfg/default, as specified in /usr/share/doc/nbd-client/README.Debian
23:57
This way all the nbd-related problems (sendsigs, not unmounting on shutdown etc) are handled by nbd itself, we don't need *any* code in ltsp at all for that anymore...
23:57
vagrantc, stgraber: Any objections on testing + commiting the dropping of poweroff -fp?
23:58
stgraber: along with the upstart scripts? (neither those are needed anymore, and they break the graceful stopping of other services...)