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


Channel log from 22 November 2008   (all times are UTC)

00:24alkisg has quit IRC
00:27hanthana has joined #ltsp
00:33vagrantc has quit IRC
00:52X0d_of_N0d has quit IRC
00:53_UsUrPeR_ has quit IRC
00:55_UsUrPeR_ has joined #ltsp
01:12X0d_of_N0d has joined #ltsp
01:40Ahmuck has quit IRC
01:41jammcq has quit IRC
01:41hesco has left #ltsp
01:41ltsppbot has quit IRC
01:47ltsppbot has joined #ltsp
02:52Q-FUNK has joined #ltsp
02:54hanthana has quit IRC
03:28plamengr has joined #ltsp
03:54hanthana has joined #ltsp
04:13alkisg has joined #ltsp
04:19Q-FUNK has quit IRC
04:25alkisg1 has joined #ltsp
04:31dirigeant has joined #ltsp
04:42alkisg has quit IRC
04:48Eghie has joined #ltsp
04:49
<Eghie>
anyone, running LTSP right now on a client with a local harddrive, which is not mounted in LTSP?
04:54hanthana has quit IRC
05:02hanthana has joined #ltsp
05:21alkisg1 has quit IRC
05:21Michiel__ has joined #ltsp
05:29Eghie has quit IRC
05:31_UsUrPeR_ has quit IRC
05:38dirigeant has quit IRC
05:40dirigeant has joined #ltsp
05:47Michiel__ is now known as Eghie
06:11alkisg has joined #ltsp
06:37Eghie has quit IRC
06:51hanthana has quit IRC
07:03Q-FUNK has joined #ltsp
07:18jammcq has joined #ltsp
07:19
<jammcq>
g'morning friends
07:19
<rjune_>
morning
07:26Q-FUNK has quit IRC
07:36Q-FUNK has joined #ltsp
07:40dirigeant has quit IRC
07:51Q-FUNK has quit IRC
07:56Q-FUNK has joined #ltsp
08:06klausade has joined #ltsp
08:13Q-FUNK has quit IRC
08:13Q-FUNK has joined #ltsp
08:14selffik has joined #ltsp
08:18mikkel has joined #ltsp
08:20alkisg has quit IRC
08:23alkisg has joined #ltsp
08:31
<mistik1>
morning guys
08:35bobby_C has joined #ltsp
08:37Q-FUNK has quit IRC
08:40
<rjune_>
mistik1: sup
09:15kharloss has joined #ltsp
09:21Q-FUNK has joined #ltsp
09:34Q-FUNK has quit IRC
09:53kharloss has quit IRC
10:07hanthana has joined #ltsp
10:36alkisg has quit IRC
10:43Egyptian[Home]1 has joined #ltsp
11:00Egyptian[Home] has quit IRC
11:10klausade has quit IRC
11:19plamengr has quit IRC
11:25chrisinajar has joined #ltsp
11:36Basti_gash has joined #ltsp
11:37Basti_gash has quit IRC
11:41alkisg has joined #ltsp
11:47Ahmuck has joined #ltsp
11:53Q-FUNK has joined #ltsp
12:14alkisg has quit IRC
12:16_UsUrPeR_ has joined #ltsp
12:18alkisg has joined #ltsp
12:21BlackDark has joined #ltsp
12:30Basti_dash has joined #ltsp
12:33Basti_dash has quit IRC
12:39_UsUrPeR_ has quit IRC
12:39_UsUrPeR_ has joined #ltsp
12:46japerry has joined #ltsp
12:48mikkel has quit IRC
13:15japerry has quit IRC
13:45johnny has quit IRC
13:45siki has joined #ltsp
13:55hanthana_ has joined #ltsp
13:56dirigeant has joined #ltsp
14:09hanthana has quit IRC
14:23kharloss has joined #ltsp
14:25hanthana_ has quit IRC
14:30chrisinajar has quit IRC
14:38japerry has joined #ltsp
14:47Egyptian[Home]1 has quit IRC
14:50japerry_cat has joined #ltsp
14:51japerry has quit IRC
14:52japerry_cat has quit IRC
14:52japerry has joined #ltsp
14:56japerry_cat has joined #ltsp
14:56ogra has quit IRC
14:57ogra has joined #ltsp
14:59japerry has quit IRC
15:03japerry_cat has quit IRC
15:03japerry has joined #ltsp
15:05Egyptian[Home] has joined #ltsp
15:08japerry has quit IRC
15:09japerry has joined #ltsp
15:11japerry has joined #ltsp
15:11japerry has quit IRC
15:12japerry has joined #ltsp
15:13Faithful has quit IRC
15:14japerry has quit IRC
15:23dirigeant has quit IRC
15:25kharloss has quit IRC
15:35alkisg1 has joined #ltsp
15:36alkisg has quit IRC
15:36alkisg1 is now known as alkisg
15:47petre has joined #ltsp
15:48
<petre>
!docs
15:48
<ltspbot>
petre: "docs" is For the most current documentation, see http://wiki.ltsp.org/twiki/bin/view/Ltsp/LtspDocumentationUpstream
16:05jstephan has joined #ltsp
16:12mathesis has quit IRC
16:16johnny has joined #ltsp
16:37captain_magnus has quit IRC
16:39selffik has quit IRC
16:43bobby_C has quit IRC
16:44captain_magnus has joined #ltsp
16:50Hyperbyte has quit IRC
16:50dirigeant has joined #ltsp
16:51alkisg has quit IRC
17:18dirigeant has quit IRC
17:25dirigeant has joined #ltsp
17:30Q-FUNK has quit IRC
17:53dirigeant has quit IRC
18:14japerry has joined #ltsp
18:29petre has quit IRC
18:41Hyperbyte has joined #ltsp
18:54siki1 has joined #ltsp
18:55siki has quit IRC
19:14siki1 has quit IRC
19:16johnny has left #ltsp
20:02japerry has quit IRC
20:05jstephan_ has joined #ltsp
20:20petre has joined #ltsp
20:21
<petre>
sbalneav, ping
20:21jstephan has quit IRC
20:27petre has quit IRC
20:37chrisinanoffice has quit IRC
20:40X0d_of_N0d has quit IRC
20:42X0d_of_N0d has joined #ltsp
20:43randra has joined #ltsp
21:06johnny has joined #ltsp
21:18Ahmuck has quit IRC
21:19Ahmuck has joined #ltsp
21:25
<sbalneav>
Evening all
21:47
<rjune_>
howdy
21:54randra has quit IRC
22:03CaScAdE^1arAway has joined #ltsp
22:07CaScAdE^FarAway has quit IRC
22:08
<jammcq>
sbalneav: Scotty !!!!!!!!!!!!!!!
22:27
<stgraber>
evening sbalneav
22:27
sbalneav: I'm working on improving screen_session so that you can switch SCREEN_XX without requiring a reboot, that's something we had quite a lot of request for ltsp-cluster but I'm not 100% sure of how to do it
22:28
for now I just moved the while loop making it recreate the environement, that works fine in most case but not all
22:28
as I'd like to start with a fresh environement everytime but as we do export I can't be sure the previous loop execution didn't set something I no longer want
22:29
easiest way I found to workaround it is to make screen_session to launch another script that'll set the environement and start the screen script, that way the environement of screen_session is always empty and as the loop is in it, every loop run the environement will be clean
22:29
I just don't know if there is another better way of doing it ...
22:32
sbalneav: http://ubuntu.pastebin.com/f3fda47d6 that's the current script I have, what I'd basically like is to make that the main() function can't change the parent environement
22:32
so if I have "export LDM_AUTLOGIN=True" somewhere in main(), the second time it's executed it won't be there anymore
22:33
the workaround I thought of is to make main() an external script instead of a function so that it'll do what I want
22:35Ahmuck has quit IRC
22:35
<sbalneav>
Hm
22:37
Not sure how we could completely clean out...
22:37
one sec.
22:37
Lemme look something up
22:39Ahmuck has joined #ltsp
22:42
<sbalneav>
Yeah, as I thought. There's no way to have a child process modify the parent.
22:42
Easiest way would be to create a function in the ltsp-common functions, called clean_env
22:43
that just does an: unset LDM_AUTOLOGIN
22:43
unset LDM_USERNAME
22:43
...
22:43
..
22:43
basically, unset all the possible ltsp variables.
22:43
<stgraber>
well, unset everything that can be set in lts.conf, that makes a lot of variables
22:43
<sbalneav>
yup :)
22:44
<stgraber>
I guess I'll just go with making main() a separate script so that it won't be able to change screen_session's environement
22:45
<sbalneav>
Making a clean function wouldn't be that hard. I could do it in about 10 minutes. Want me to do it.
22:45
?
22:45
might be useful in other places eventually too.
22:46
<stgraber>
well, do we have a good way of determining all non-standard environement variables (as in all that aren't set when you start a shell) ?
22:46
I don't want a static list as you can't now what custom variables will be used for some user-done screen.d scripts or ldm rc.d scripts
22:48
<jammcq>
seems like it would be nice to have a documented standard naming convention for variables. a well-behaved screen script would follow that standard, and scotty's clean-up script would have an easy job
22:48
<sbalneav>
Hmm, that's true.
22:49
yeah, I was just thinking. All ldm variables start with LDM_
22:49
if all the other variables started with LTSP_ or the like, then it would be easy to tell.
22:49
<jammcq>
well, it's never too late to start
22:50
<sbalneav>
:)
22:50
<Ryan52>
well, some of the gtkgreet variables also start with LDM_. they should be GTKGREET_ instead. especially since someday the Qt greeter might be ready :P
22:51
anyway, that's irrelavant to this conversation. *goes back to whatever he was doing*
22:51
<jammcq>
the clean-up script could have a plug-in capability. if you install a QT greeter, it should include a QT cleanup plug-in
22:51
<sbalneav>
would the gtkgreeter variables be general enough to use for a Qt greeter? Could we just use GREETER_THEME, etc...
22:51
<jammcq>
LDM could include a clean-up plug in as well
23:03spud1 has joined #ltsp
23:04
<spud1>
Hi. What became of k12os.org?
23:05
<stgraber>
sbalneav: what about something like: unset $(/usr/bin/env | egrep '^(\w+)=(.*)$' | \ egrep -vw 'PWD|USER|LANG' | /usr/bin/cut -d= -f1);
23:05
sbalneav: so we'd do it the other way around, specifying what we want to keep
23:07
<jammcq>
stgraber: so when a user logs out, are you going to re-parse the lts.conf file before allowing the user to log in?
23:07
<stgraber>
jammcq: basically what we want for ltsp-cluster is: you logout, X stops, you come back to screen_session, it gets what screen script needs to be executed and the rest of the environement and start the screen script
23:08
so it's like you're rebooting the computer as far as the configuration is concerned
23:08
you can basically change everything and it'll be updated at logout
23:08
<jammcq>
yes, but..... that's how LTSP used to work, back around LTSP-4.0. we changed it around because we wanted to speed up the time to re-login
23:09
<stgraber>
one of the request we have is in case where the SAN fails, so the session will crash, they will reload the config with LDM_AUTLOGIN=True and load an autologin session that doesn't have network mounted /home
23:09
currently they need to reboot 3k thin clients to do that
23:10
(taking like an hour to do even when remotely rebooting because of the dhcp/tftp/nbd servers being overloaded by the whole network rebooting)
23:10
<jammcq>
how often does the SAN fail?
23:10
<stgraber>
often enough for them to ask for a way to workaround it :)
23:10
<jammcq>
seems like you'll be penalizing EVERYBODY for the off chance that a SAN would fail, which i'd imagine would be a rare occurrance
23:10
<rjune_>
that's scary
23:11
<jammcq>
seems like they should fix their SAN infrastructure
23:11
<stgraber>
but same goes if the LDAP server goes done, ...
23:11
<rjune_>
jammcq: I'm with you
23:11
<stgraber>
they are also working on that with the guys at Novell
23:12
<jammcq>
but there's some optimization that's been done to speed up the logout->login experience and you'll be removing that
23:12
<stgraber>
we are asked to provide high availability and we don't choose what SAN technology they are using so we need HA even in degrated mode which is why we spent time implementing our autologin
23:13
<jammcq>
what starts the screen scripts now?
23:13
<stgraber>
jammcq: depends, we can call the reset environement only when /etc/ltsp/getltscfg-cluster.conf is present, so only affecting the ltsp-cluster setup
23:13
jammcq: screen_session does
23:14
<jammcq>
It's been a long time since i've looked at any of the code, but I think we used to start the screen scripts from init
23:14
if you detect an error in a screen script, you could drop out of the script, which would cause init to respawn, giving you a fresh environment
23:15
or maybe it was init->screen_session->screen_script
23:15
<stgraber>
nope, currently the loop is in screen_session, that's what's respawning the screen script
23:15
<jammcq>
so screen_script could return something back to screen_session, and screen_session could be told to die, which should then be respawned by init
23:15
or upstart
23:15
again, giving you a fresh environment
23:16
a simple logout shouldn't cause screen_session to terminate
23:16
<stgraber>
yeah, having it as a /etc/event.d script is what I'd like to have at the end but not until every distribution has it
23:16
<jammcq>
so it would still be fast
23:17
<stgraber>
I guess that for now, I just will reset the environement when ltsp-cluster is detected using something similar to the one-liner I copy/pasted before
23:18
so there won't be any performance change for regular LTSP and the ltsp-cluster users will be happy. If we then want it for regular ltsp too, we'll just need to make it so it use the same function and re-downloads lts.conf
23:18
sbalneav: ^ does that make sense ?
23:18
<jammcq>
well, that way, you have what you want, and if it doesn't cause a performance problem, it's a simple change to make it that way for everybody
23:20
<sbalneav>
Yeah, makes sense.
23:29izad has joined #ltsp
23:29|Paradox| has quit IRC
23:30izad is now known as |Paradox|
23:31
<spud1>
So, is it stiill feasable to do a OpenMosix type cluster with k12ltsp?
23:31
<jammcq>
does Openmosix still exist?
23:32
<spud1>
discontinued since March
23:32
<jammcq>
hmm, I thought it was gone years ago
23:32
is there something to replace it?
23:33
<spud1>
Maybe, but I think you can still get it
23:34
<jammcq>
well, because LTSP is now based on distro packages, including the standard kernel, it should be fairly straightforward to use it for clustering
23:35
if you can make normal disk-based linux work, doing it with ltsp should be almost the same
23:36
<spud1>
Seems to make sence to bring in all those cpus
23:36
You could use a lower end server
23:37
<jammcq>
well, are you talking about making a compute farm? or maybe you just want to run each users apps locally, on their thin client
23:38
<spud1>
Don't know. A farm, maybe.
23:39
<jammcq>
stgraber is working on ltsp-cluster, maybe that's what you need
23:39
<spud1>
Thanks I'll look.
23:47
<stgraber>
ok, building a new ltsp chroot with my changes to screen_session to make sure LTSP isn't affected at all. If it's fine I'll build some test packages to test the ltsp-cluster part, I hope to be able to tag a new LTSP tomorrow after I did my tests.
23:47
at this point ltsp-cluster will be entirely merged for its client part (chroot)
23:50alkisg has joined #ltsp