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


Channel log from 1 September 2007   (all times are UTC)

00:04
<sbalneav>
Night all
00:04sbalneav has quit IRC
00:42snorp_ has quit IRC
00:57highvolt1ge has quit IRC
01:13tux_440volt has joined #ltsp
01:16zodman has quit IRC
01:21cyberorg has quit IRC
01:23cyberorg has joined #ltsp
01:38zodman has joined #ltsp
01:39vagrantc has joined #ltsp
01:40tux_440volt has quit IRC
02:08bobby_C has joined #ltsp
02:52vagrantc has quit IRC
03:13zodman has quit IRC
03:30bobby_C has quit IRC
03:46K_O-Gnom has joined #ltsp
03:58mcfloppy has quit IRC
04:36ace_suares has quit IRC
04:36ace_suares has joined #ltsp
06:58ltsppbot has quit IRC
07:01cliebow has joined #ltsp
07:08cyberorg has quit IRC
07:11K_O-Gnom has quit IRC
07:16K_O-Gnom has joined #ltsp
07:47Q-FUNK has joined #ltsp
08:36plamengr has joined #ltsp
08:38ltsp-sid has joined #ltsp
08:46daya has joined #ltsp
08:58ltsppbot has joined #ltsp
09:11mohamed_ has joined #ltsp
09:11sbalneav has joined #ltsp
09:12
<sbalneav>
Morning all
09:12
<Faithful1>
scotty!
09:16
<sbalneav>
hey faithful
09:43muh2000 has quit IRC
10:11ari_stress has joined #ltsp
10:14bobby_C has joined #ltsp
10:21ltsppbot has quit IRC
10:25ltsppbot has joined #ltsp
10:33daya has quit IRC
10:35Egyptian[Home] has joined #ltsp
10:36jim_k has joined #ltsp
10:37
<jim_k>
anyone in here that follows the edubuntu-users list?
10:38
sbalneav....you around?
10:40
I need help troubleshooting a hard freeze on clients when opening graphic intense websites
10:41
Or at least a confirmation that others have client freezes when opening a site such as http://www.carteretcountyschools.org/bms/teacherwebs/sdavenport/artgallery6.htm
10:43vagrantc has joined #ltsp
10:43
<jim_k>
I must have joined during lunch break :-)
10:45
vagrantc.....are you available?
10:48
I am wondering if anyone can see my posts?
10:49
help needed troubleshooting client freezes with graphic intense websites
10:49jim_k has quit IRC
10:50jim_k has joined #ltsp
10:50
<vagrantc>
jim_k: see your posts ?
10:50
posts where?
10:51
<jim_k>
here....just did see anyone responding
10:51
wasn't sure if I was logged in right
10:51
everyone must just be busy...thanks for the reply
10:51
<vagrantc>
just ask your questions, be patient, and someone will respond when they can
10:52jammcq has joined #ltsp
10:52
<sbalneav>
jim_k:
10:52
<vagrantc>
jammcq: mornin
10:53
<sbalneav>
I was looking at things earlier.
10:53
<jim_k>
Okay....I need help troubleshooting clients freezing hard when accessing graphic intense websites
10:53
<jammcq>
hey kids
10:53
<sbalneav>
This is kind of an old problem.
10:53
hey jammcq
10:53
<jammcq>
jim_k: hey, still at it, eh?
10:53
<jim_k>
If I hit a site such as http://www.carteretcountyschools.org/bms/teacherwebs/sdavenport/artgallery6.htm the client freezes hard
10:53
jammcq...yep :-(
10:53
I just posted the results of running free on the edubuntu-users list
10:54
sbalneav....old problem huh, crap, this is making things almost unusable here
10:54
first week of school after convincing everyone to finally ditch all the old macs and go linux
10:55
now students are constantly freezing while browsing the web
10:55
even worse, when they freeze processes are left over, making it so they can't log back in again without errors
10:55
so I'm here all weekend until I figure out how to make this work on Tuesday :-(
10:56
<ari_stress>
sorry to hear that jim_k
10:56
<jammcq>
jim_k: someone needs to slap that website creator. using width and height to scale a big image is just wrong
10:56
<sbalneav>
There's no way to really fix it, but there is a way to help it somewhat.
10:57
<ari_stress>
will adding more ram help?
10:57
<jim_k>
jammcq..yeah, but our art class doesn't really care when they just search for outrageously large paintings
10:57
<jammcq>
ari_stress: yes, adding more ram helps
10:57
<jim_k>
ari_stress...adding ram helps
10:57
<jammcq>
jim_k: crank swap up even higher
10:57
<sbalneav>
The problem is that Firefox simply keeps allocating pixmap space on the thin client, until the x server basically exhausts ram completely
10:57
<jim_k>
but even with 512MB on the client from 128MB I still crash on the above listed site
10:58
jammcq...is there a problem with having too large a swap? I'll set it to 1GB if need be :-)
10:58
<sbalneav>
however, jammcq and I came up with a solution, which might help here, and I've looked up.
10:58
<jim_k>
sbalneav....all ears
10:58
<jammcq>
well, the problem is, you can't fit 10 lbs of crap in a 5lb bushel basket
10:58
<ari_stress>
i'm opening the site now, and it's ridiculously heavy
10:58
<jammcq>
firefox sends the images to the Xserver to be displayed
10:59
<jim_k>
is there a way to tell the client to quite loading the crap into the basket before it crashes?
10:59
<sbalneav>
ok, you're using ldm, correct?
10:59
<jammcq>
jim_k: you can tell the Xserver to only allocate some percentage of total available ram.
10:59
<jim_k>
sbalneav....yep....ldm...tried with and without DIRECTX hack
10:59
<jammcq>
but, what happens is the next thing that asks the Xserver for ram will get denied
10:59
<jim_k>
jammcq...will that help it to not crash
11:00
<sbalneav>
ok, so, lets go to /opt/ltsp/i386/usr/lib/screen.d/ldm, and edit it.
11:00
<jammcq>
unfortunately, Xclients aren't usually written to deal with the denial of ram, so the client crashes
11:00
<jim_k>
at this point I'd be okay with the site not loading, and the client not crashing
11:00
<jammcq>
when it's firefox, then firefox dies. prolly not a big deal. but.... what if it's Gnome that makes the request. For instance, you click a menu, it causes a memory request. then Gnome crashes
11:01
<jim_k>
no screen.d directory
11:01
<sbalneav>
sorry, /opt/ltsp/i386/usr/lib/ltsp/screen.d
11:01
<jim_k>
firefox dies...okay with that, gnome dies and goes to login screen...okay with that too, just client freezing bad
11:02
okay ldm open for editing
11:02
well I guess I'm not "okay" with the crashing, but it would be a stop forward
11:03
<sbalneav>
ok, just before the line that says: ldm ..., down at the bottom?
11:03
<ltsppbot>
"sbalneav" pasted "Additions to ldm" (16 lines) at http://pastebot.ltsp.org/298
11:04
<sbalneav>
paste that in.
11:04
then, add the following line to your lts.conf file:
11:04
<jim_k>
ahh, shoot, I don't see a line that starts with ldm
11:05
<sbalneav>
it's at the bottom
11:05
<jammcq>
sbalneav: could you post a few lines from that file, so i can see it?
11:05jammcq has quit IRC
11:05
<sbalneav>
which, the ldm screen script?
11:05
<jim_k>
vim /opt/ltsp/i386/usr/lib/ltsp/screen.d/ldm last line is exec ldm vt$ttynum :$displaynum
11:05
<sbalneav>
that's the one.
11:06
paste what I posted just before that line.
11:06jammcq has joined #ltsp
11:06
<jammcq>
sorry, we have a new dog that walked across my keyboard and closed my irc window
11:06
<jim_k>
okay, I pasted the modified text
11:07
<sbalneav>
New dog?!?! \o/
11:07
<jim_k>
jammcq....:-)
11:07
<jammcq>
little 2.25lb chahuahua
11:07
<sbalneav>
bs
11:07
:)
11:08
ok, so now, add the following line to your [defaults] section in lts.conf
11:08
XRAMPERC=75
11:08
<jim_k>
jammcq....you get one of those purses like paris hilton to put the dog in :-)
11:08
<jammcq>
jim_k: oh yeah, the whole thing
11:08
<jim_k>
lts.conf modified
11:09
<jammcq>
it's amazing the energy a little 2 pound dog can have
11:09
<jim_k>
my wifes parents have little dogs, energy never stops
11:09
<sbalneav>
now, reboot a client, and lets make sure it boots.
11:09
<jammcq>
my fear is that I'll step on the little thing
11:10
<jim_k>
jammcq...that wouldn't be good :-)
11:10
client booting okay so far
11:11
so this hack should keep X from using more than 75% of available RAM?
11:11
<sbalneav>
jammcq: You're serious?
11:11
<jammcq>
sbalneav: about what?
11:11
<sbalneav>
You've got a chahuahua?
11:11
<jim_k>
client booted up...I'll log in
11:11
<jammcq>
yeah
11:11
a REAL cha hua hua
11:11
black
11:11
<sbalneav>
LOL, thought you were joking.
11:11
<jammcq>
pff, dead serious
11:12
<jim_k>
crap forgot to kill old processes, rebooting
11:14
booted up...logged in successfully
11:14
Should I try crashing it :-)
11:15
<sbalneav>
I'd say, yes.
11:15
<jim_k>
here goes nothing
11:16
well...the second I go to load one of those pages firefox crashes and I'm back to my desktop
11:16
<jammcq>
there ya go
11:16
<jim_k>
it doesn't look like it even tries to use nbd now
11:17
I play with this and see how it goes
11:17
I assume I can play with the percentage used in XRAMPERC
11:17
<jammcq>
yeah, even try 95%
11:17
<sbalneav>
yup.
11:17
<jammcq>
but with that website, yer gonna need more than that
11:17
<jim_k>
thanks guys
11:17
<jammcq>
so try increasing swap
11:17
<sbalneav>
That's what will happen, firefox will just die.
11:17
<jammcq>
well, *something* will die
11:18
<jim_k>
will increasing swap help with this ?
11:18
<jammcq>
you HOPE it's firefox
11:18
yes
11:18
umm
11:18
<sbalneav>
I'll add this in for gutsy.
11:18
<jim_k>
is there a problem with setting a ridiculous swap size?
11:18
<jammcq>
sbalneav: does that thing look at TOTAL virtual ram? or just physical ram ?
11:18* jammcq doesn't remember
11:18* vagrantc proposes X_* for all X related variables
11:18
<sbalneav>
total, I beleive
11:18
ok, x_ramperc then.
11:19
<jammcq>
yeah, total
11:19Faithful1 has quit IRC
11:19
<jim_k>
when I run free it doesn't show any nbd swap getting used before crashing firefox
11:19
<vagrantc>
ramperc sounds like something you'd have on a castle
11:19
<sbalneav>
lol
11:19
<vagrantc>
man the ramperc's!
11:19
<jammcq>
jim_k: how much physical ram do you have?
11:19
<sbalneav>
rampercentage is what it's supposed to stand for.
11:20
<jim_k>
physical is 128MB
11:20
I have a 512MB stick for testing
11:20
but the 128 is in right now
11:20
<sbalneav>
This will help it to "fail" a bit better.
11:21
<jim_k>
say I increase /etc/ltsp/nbdswapd.conf to SIZE=1024....should things run okay?
11:21
<sbalneav>
No idea
11:21
<jammcq>
jim_k: TRY IT
11:21
<sbalneav>
yeah, dude :)
11:21
<jammcq>
you won't break anything
11:21
at least not permanently
11:22
<jim_k>
will the ramperc option look at total virtual ram, then if there isn't enough just crash the app? It seems like that is what is happening
11:22
rebooting with SIZE=1024
11:22
<vagrantc>
if you've got 1Gb of extra disk space on your server for each and every client ....
11:22
<jammcq>
jim+kyeah
11:23
damned dog :)
11:23
jim_k: yes
11:23
<jim_k>
hard drive space is way cheaper and easier than 108 512MB DDR 400 RAM sticks?
11:23
<sbalneav>
firefox is making memory requests. if it makes one, and the x server tells it "no, I can't allocate that", then firefox dies.
11:23
<jammcq>
it's hard to type when a dog is walking across the back of your hands
11:23
<jim_k>
jammcq...LOL
11:23
<jammcq>
sure would be nice if firefox would fallback gracefully
11:24
<jim_k>
an error saying not enough ram for this site then okay to return to last site would be awesome
11:24
<jammcq>
or just show images as broken images or something
11:24
<sbalneav>
jim_k: That's a firefox bug.
11:24
<jim_k>
anything better than "poof" what happened....stupid linux :-
11:24
<jammcq>
but it's not just images. it's fonts and other things too
11:24
<sbalneav>
file one with them
11:25
jim_k: Not linux's fault.
11:25
it's firefoxes fault.
11:25
<jim_k>
I know....but that is what the staff and students will say...not me :-)
11:25
me knowing what is happening doesn't help their perception
11:25
all they know is the old mac opened up the site, and the new system doesn't
11:25
<jammcq>
certainly jim_k is on the hook to deliver a working solution. it doesn't really matter to the end user what the details are. if it doesn't work.... it doesn't work
11:26
<sbalneav>
So, then stick with the old macs.
11:26
<jammcq>
add more ram
11:26
how's the swap=1024 working?
11:26
<jim_k>
sbalneav: I hate macs though :-)
11:26
<vagrantc>
or stick hard drives in your thin clients with huge swap partitions
11:26
<jim_k>
jammcq: loading now
11:28
massive swap was working, switched to screen1 to check free, swap getting used, switched back and hung
11:28
I'll try rebooting and loading without the switch to check free
11:28
<vagrantc>
some things can't be swapped out ...
11:28
unfortunately
11:28
<jim_k>
vagrantc: small notebook drives might be close to the same cost of 512MB ram sticks...hhhmmmm
11:29
<jammcq>
can you put more than 512mb in those thin clients?
11:29
<jim_k>
jammcq: they'll handle a single 2GB stick....but then cost gets to be an issue as well
11:30
but if things worked for $8,000 more, compared to not working for $8,000 less, might not be all that bad :-)
11:30
<sbalneav>
The $8k you'll make up in reduced maintenance costs.
11:30
<vagrantc>
jim_k: what was the motivation to switch ?
11:31muh2000 has joined #ltsp
11:31
<jim_k>
vagrantc: cost was part of the motivation (with admin), for me software,administration,and linux freedom is the reason
11:32
<sbalneav>
vagrantc: ok, pushing new rev.
11:32
ok, I'm off to the beach for the day.
11:32
I'm afk.
11:32
<vagrantc>
sbalneav: new rev of what where ?
11:32
<jammcq>
beach ?
11:32
in Canada ?
11:32
<sbalneav>
vagrantc: ltsp
11:33
a bugfix, and X_RAMPERC
11:33
off
11:34
<jim_k>
on phone right now....
11:38
thanks scott
11:56
<klausade>
how is support for audio-cd in ltsp5?
11:57
<jim_k>
opera works perfectly without crashing with no nbd or xramper hacks
11:57
<jammcq>
well
11:57
there ya go
11:58
<jim_k>
I am going to file a bug report with firefox
11:58
<jammcq>
opera must be better at telling the xserver to free the memory that was requested
11:58
<jim_k>
I'll also look into a hack to disable image caching in firefox
11:58
<jammcq>
jim_k: I'm pretty sure there's already a bug report for that. I think I filed one years ago
11:58
but search for it
12:02fernando1 has quit IRC
12:03
<vagrantc>
jim_k: image caching in firefox is just on disk, as far as i know
12:04
<jim_k>
vagrantc: does that mean it shouldn't be happening on a thin client
12:04
<vagrantc>
jim_k: yes, it'll be server-side
12:05
<jammcq>
it's not necessarily image caching. when firefox wants to display an image, it sends the image to the Xserver. The xserver allocates memory for it. when the browser is done with that page, it should ask the Xserver to destroy the image object, releasing the memory
12:06
firefox doesn't do that very well
12:06
<jim_k>
there has to be a way to hack firefox and keep it from doing this, opera seems to have it figured out
12:06
<vagrantc>
jim_k: might try other browsers like epiphany or konqueror and see how they handle the sites
12:06
<jammcq>
when you load a long page, like the link you pasted, there's lots of images. not all are visible on the screen at the same time
12:07
<jim_k>
not only does opera work, but it works way faster
12:07
<jammcq>
firefox sends them ALL to the Xserver in anticipation of you scrolling the page to make them visible
12:07
<jim_k>
vagrantc: turns out opera works flawlessly without the need for xramperc or nbdwap hacks
12:07
<jammcq>
jim_k: in firefox, go to 'about:config'
12:08
<vagrantc>
yes, but opera isn't free software, whereas epiphany and konqueror are
12:08
<jim_k>
firefox acting this way seems pointless, opera isn't doing it and it is faster
12:08
<jammcq>
look at the entry called: browser.cache.memory.enable
12:08
and try setting that to 'false'
12:08
see if it makes a difference
12:08
<jim_k>
vagrantc: epiphany, galeon, and firefox all act the same way
12:08
vagrantc: not saying opera is better, just that it works
12:08
jammcq: I'll try that
12:09
<vagrantc>
jim_k: try what jammcq says, or try konqueror, which uses an entirely different rendering engine, then
12:09
<jammcq>
jim_k: what's that URL again?
12:09
<jim_k>
I'll try konquerer as wel
12:09
jammcq: http://www.carteretcountyschools.org/bms/teacherwebs/sdavenport/artgallery6.htm
12:10fernando1 has joined #ltsp
12:11
<jammcq>
hmm, that setting doesn't seem to matter
12:11
<jim_k>
how about browser.cache.disk.capacity set lower than 50000?
12:12
<vagrantc>
disk has nothing to do with it
12:12
<jammcq>
yeah, that's server side
12:12
<jim_k>
yep...still froze :-)
12:13
was on the phone with Robert Arkiletian....he has a version of TeacherTool almost ready for Edubuntu
12:13
<jammcq>
so, maybe your solution for next week is Opera
12:14
<jim_k>
for now...yes...definitely opera...but I sure would like to find a solution to get firefox to work, I'm sure others would appreciate it
12:14
especially since firefox is the default in edubuntu
12:15fernando1 has quit IRC
12:15
<jim_k>
clients hard crashing with random web searching just isn't a good thing, especially when it doesn't have to happen
12:16
firefox will be removed for now on Tuesday and Opera will be the new default.....thanks again sbalnaev,jammcq,vagrantc for the help
12:16
<vagrantc>
people have been talking about these sorts of bugs in firefox for years...
12:16
<jim_k>
this is why linux and ltsp rock
12:17
less talk...more action :-)
12:17
<vagrantc>
for whatever reason, it's not a priority to get them fixed
12:17
<jim_k>
so the firefox developers don't see ltsp as a priority huh?
12:17
<Q-FUNK>
I really hope that gnome adopts webkit for the next iteration of epiphany.
12:17
<vagrantc>
maybe someone could get it fixed in iceweasel, and then iceweasel puts the pressure on
12:17
<jammcq>
jim_k: no, they see Windows as a priority
12:18
<jim_k>
jammcq: fools
12:18
<rcy>
jim_k: I met Robert A. at freegeek the other night... looking forward to seeing his TeacherTool on our debian ltsp soon
12:18
<Q-FUNK>
jammcq: if even that. to quote sabdfl: the problem is that developers are only willing to scratch THEIR itch.
12:18
<jim_k>
rcy: TeacherTool is awesome
12:18
<vagrantc>
rcy: you from freegeek vancouver ?
12:19
<rcy>
yeah, i actually havent seen it yet, but how robert described it, i was really stoked
12:19
vagrantc: yes
12:19
<vagrantc>
i gotta get up there sometime
12:19
<rcy>
vagrantc: yes! david and I wanted to come down to portland again soon, but we cant find the time to get away
12:19
<jim_k>
I'm going to send Robert A. Scott's client shutdown hacks to see if he can incorporate a way to shut down clients in TeacherTool
12:20
<vagrantc>
rcy: so you're using debian's ltsp ? from etch, or with the backports ?
12:20
<Q-FUNK>
jammcq: my perception is that even windows is not a priority to them. geting their pet features is.
12:20
<rcy>
vagrantc: debian etch, ltsp 4.2. i really had know idea what i was doing when i installed it... differences in versions etc, but its going pretty well, all in all
12:20* vagrantc pouts
12:21
<jammcq>
vagrantc: I thought you were a fedora dev now
12:21
<vagrantc>
jammcq: pfft
12:21
<rcy>
I want to find some servers to experiment with 5, and *ubuntu
12:21
though, robert sorta scared me away from 5
12:21
<mistik1>
hey fellas
12:22
<vagrantc>
rcy: yeah, well, you've got one of the most active LTSP5 developers in the freegeek community... :)
12:22
<jim_k>
rcy: I think ltsp5 is awesome, but different from 4.2....Robert has tweaked on 4.2 for years and the change is a little scary for him I think
12:23
<rcy>
the biggest thing he said was that the way that 5 does the communication with the clients (ie not xdmcp as I understand it) introduces a lot of latency in the user interface
12:23
you need much much bigger thinclients
12:23
<vagrantc>
that's optional now
12:24
it's still the default, but turning it off is a simple configuration option
12:24
<rcy>
right, which i think he also mentioned...
12:24
<mistik1>
jammcq: WOuld it be posible to open that imap port to my current IP sir?
12:24
<jim_k>
rcy: now with the option of LDM_DIRECTX in ldm I think it is just as responsive as 4.2
12:25
<rcy>
what is the best thing to install for ltsp 5? ubuntu, edubuntu, debian?
12:25
<jammcq>
mistik1: sure, just give me the details
12:25
<vagrantc>
rcy: edubuntu includes it as part of the default server install ...
12:25
<rcy>
vagrantc: who is the LTSP 5 dev in vancouver?
12:25
<jim_k>
rcy: I have been very happy with the ease of Edubuntu
12:25
<mistik1>
jammcq: 67.82.199.211
12:25
<rcy>
edubuntu, ok... ill test drive that on another server one of these days
12:26
<vagrantc>
rcy: sorry, i was referring to myself being an active LTSP5 developer in the freegeek community ... not vancouver
12:26
<jim_k>
rcy: I think you'll like what you find :-)
12:26
<jammcq>
mistik1: ok, give that a try
12:26
<vagrantc>
rcy: but if you prefer debian or plain ubuntu, it's not difficult to set up
12:26
<rcy>
vagrantc: ah right, you clearly said that... i misread
12:26
<vagrantc>
rcy: easy to do ...
12:26
talk to y'all later
12:27
gotta go get thrown around some :)
12:27vagrantc has quit IRC
12:27
<mistik1>
jammcq: says LOGIN disabled by server
12:27
<jammcq>
huh?
12:27
<mistik1>
lemme go over what I had setup in the client
12:28
<rcy>
vagrantc: i dont prefer one over the other really (I'm a bsd guy), so I want to install the best deb based system for maintaining the lab and providing what we need for our users
12:28
<jammcq>
mistik1: I see you trying
12:28
rcy: vagrant has left the building
12:29
<rcy>
the biggest holes in the current system is lack of get local disk access (usb keys, cdrom/rw) and better management of guest accounts
12:29
<mistik1>
jammcq: This stupid client keeps looping
12:29
<rcy>
jammcq: i should do the same
12:30ari_stress has quit IRC
12:35
<mistik1>
jammcq: close the port again for a minute
12:35
so I can config kmail without it going nuts
12:36
<jammcq>
ok, closed
12:36monteslu has quit IRC
12:46tux_440volt has joined #ltsp
12:52subhodip_ has joined #ltsp
12:52fernando1 has joined #ltsp
12:59ltsp-sid has quit IRC
13:07tux_440volt has quit IRC
13:09
<jim_k>
it appears the olpc guys ran into the pixmap problem in firefox as well...looking to see if they found a fix
13:19indradg has joined #ltsp
13:19rcy is now known as legacy_nick
13:20subhodip_ has quit IRC
13:32jim_k has quit IRC
13:45monteslu has joined #ltsp
13:57kaminski-ltsp-br has joined #ltsp
13:58legacy_nick is now known as rcy
14:06tux_440volt has joined #ltsp
14:07plamengr has quit IRC
14:31vagrantc has joined #ltsp
14:57tux_440volt has quit IRC
15:24multik has joined #ltsp
15:25
<multik>
hi there sbalneav :)
15:25
Q-FUNK, greetings :)
15:41kaminski-ltsp-br has quit IRC
15:47joebob777as7 has joined #ltsp
15:51indradg_ has joined #ltsp
16:03multik has quit IRC
16:05mohamed_ has quit IRC
16:05bobby_C has quit IRC
16:05bobby_C has joined #ltsp
16:11indradg__ has joined #ltsp
16:11indradg has quit IRC
16:13Q-FUNK has quit IRC
16:28bobby_C has quit IRC
16:29indradg_ has quit IRC
16:29indradg__ has quit IRC
16:31indradg has joined #ltsp
16:31ltsp-sid has joined #ltsp
16:35vagrantc has quit IRC
16:35indradg_ has joined #ltsp
16:43cliebow_ has joined #ltsp
16:43cliebow has quit IRC
16:55indradg has quit IRC
17:08joebob989823 has joined #ltsp
17:08joebob777as7 has quit IRC
17:15loather-work has quit IRC
17:17loather-work has joined #ltsp
17:23muh2000 has quit IRC
17:35K_O-Gnom has quit IRC
17:37ltsp-sid has quit IRC
17:49vagrantc has joined #ltsp
18:14joebob989823 has quit IRC
19:00captain_magnus has quit IRC
19:01captain_magnus has joined #ltsp
19:07spectra has joined #ltsp
20:19ccjoe has joined #ltsp
20:29ari_stress has joined #ltsp
20:30
<ccjoe>
I am trying to get LTSP 5 working on Ubuntu 7.04, the client comes up to the login screen but I am unable to login. It looks like it is attempting to login to the client instead of the server. Any ideas?
20:30
<ari_stress>
morning
21:20J45p3r has joined #ltsp
21:21
<sbalneav>
ccjoe: You probably need to do an ltsp-update-sshkeys
21:22
<ccjoe>
ok, I'll try it
21:29
<vagrantc>
sbalneav: please merge my switch back to using the debian logo in the debian theme: http://llama.freegeek.org/~vagrant/bzr/ltsp/ubuntu/vagrant-gutsy
21:33
<sbalneav>
vagrantc: Done, pushing
21:34
vagrantc: Pushed.
21:39
<vagrantc>
sbalneav: thanks. i like when i'm booting a debian LTSP for the debian logo to show up :) (vs. edubuntu)
21:39
<sbalneav>
Hmm
21:39
I'm trying to get the workstation to come up with a proper hostname.
21:40
got a sec?
21:40
<vagrantc>
sure
21:40
is /opt/ltsp/i386/etc/hostname present ?
21:40
<sbalneav>
yes
21:40
<vagrantc>
what's in there ?
21:40
<sbalneav>
edubuntu
21:40
I've got get-lease-hostnames true; in my dhcpd.conf
21:41
<vagrantc>
remove the hostname file ... maybe that'll work
21:41
<sbalneav>
and I've got all the hostnamed defined in /etc/hosts
21:41
hm
21:41
<vagrantc>
or: echo > hostname
21:42
if there's anything in /etc/hostname, debian (and presumably ubuntu) will use that as the hostname.
21:42
<sbalneav>
ok
21:42
I'll try moving it.
21:42
<vagrantc>
there are two long-standing bugs at http://bugs.debian.org/src:ltsp
21:42
i haven't figured out a good way to fix it.
21:43
without re-introducing the problem that hostname gets set to "(none)" if there's no DNS
21:43* vagrantc sighs
21:44
<vagrantc>
i was hoping i'd get a chance to test nbdroot and stuff today
21:44J45p3r has quit IRC
21:44
<sbalneav>
If hostname gets set to "(none)", does that bugger something up, or just look bad?
21:44
<vagrantc>
it just looks stupid
21:44
<sbalneav>
That's easy, then
21:45
<vagrantc>
we'd have to have a hook somewhere extremely early in the boot process
21:45
<sbalneav>
we right a plugin to move it out of the way, and then in the setup, just test if it's (none), and if it is, set it to something sensible.
21:45
<vagrantc>
as rcS.d/S02hostname.sh is what sets it
21:46
yeah, but do we want a whole new init script just to handle that ?
21:46muh2000 has joined #ltsp
21:46
<sbalneav>
couldn't we just do it later in the process, in the ltsp-client-core script?
21:47
<vagrantc>
then for part of your boot process, you're running as one hostname and switch to another ...
21:47
so any other init scripts depending on host name might behave strangely
21:48
<sbalneav>
ok, how about putting something in the initramfs to take the passed hostname, and overwrite /etc/hostname?
21:48
yeah, if I move /etc/hostname to /etc/hostname.old, name gets set properly
21:48
<vagrantc>
that's what ogra proposed
21:49
actually, if the initramfs just sets hostname and /etc/hostname is removed or empty, that should also work.
21:53
<sbalneav>
Is that something we could do, or would it mean modifying initramfs-tools itself?
21:55
<vagrantc>
initramfs-tools is very flexible ..
21:55
i suspect we just need the hostname binary in the initramfs
21:56
<sbalneav>
does /proc get mounted in the initramfs?
21:56
just cat /proc/sys/kernel/hostname?
21:57
must be, we check /proc/meminfo
22:01
vagrantc: The initramfs is the one grey area I have, how would we create a plugin for hostname?
22:02
All we'd have to do is read the hostname from /proc/sys... into a var, and if it's test -n, then mv /etc/hostname to /etc/hostname.old
22:02
we'd have to do it at the bottom of, or after, nbd root.
22:03
something in the nfs-bottom series of scripts?
22:06
<vagrantc>
if we can get a sane hostname, let's just avoid /etc/hostname altogether and remove it
22:06
or zero it out
22:08
i can't figure out why initramfs-tools isn't including the ltsp-nbd script
22:09
ah, initramfs-tools uses shell variable names for handling dependencies. Notice that ‘-’ is not a valid shell variable name and thus should not be used as script filename.
22:15
<sbalneav>
So are you doing something on this now?
22:16
<vagrantc>
on what ?
22:16
i'm trying to get an upload for debian that's synced with ubuntu
22:18LTSP has joined #ltsp
22:34Faithful has quit IRC
22:35Faithful has joined #ltsp
22:35
<vagrantc>
seems like it's not waiting for /dev/nbd0 to work
22:35
<jammcq>
why not just set the hostname in dhcpd.conf ?
22:36
<vagrantc>
before trying all sorts of other mounts
22:37
<sbalneav>
jammcq: It is being set there, the problem is, by default, there's an /etc/hostname file in the chroot
22:37
<vagrantc>
jammcq: go for it
22:37
<sbalneav>
and what ends up happening is, if /etc/hostname is there, it ends up overriding the one set by the dhcp info.
22:38
<jammcq>
so make it not be there
22:38
<sbalneav>
Well, the problem is, though, that if dhcpd DOESN'T set one, then you;d like to have a sane default hostname.
22:38
<jammcq>
it looks like the default one contains the same name as the server
22:39
<sbalneav>
So, all we need to do is just check /proc/sys/kernel/hostname in the chroot
22:39
<jammcq>
which doesn't seem sane to me
22:39
<vagrantc>
ok, so nbd-client returns a failure almost instantly.... no nbdroot for me.
22:39
<sbalneav>
if it's been set (by the dhcpd), then just remove /etc/hostname.
22:40
<jammcq>
otherwise default to the hostname of the server?
22:40
<vagrantc>
that'll work for unionfs ...
22:40
in debian, we feed a default of "ltsp" for the hostname
22:40
<sbalneav>
jammcq: Well, that file just gets filled in on the chroot build.
22:40
<jammcq>
sbalneav: yeah, I understand that. but it seems if you fill it in with the servers hostname, that would be wrong
22:41
<vagrantc>
that's why i fill it in with "ltsp"
22:41
<sbalneav>
Well, at the time you're building the chroot, it's the only hostname you've got.
22:42
At least if it's the server's hostname, you know which root you've mounted :)
22:42
<vagrantc>
my thoughts: no /etc/hostname at all, initramfs looks for a hostname, and if it doesn't find one, sets it to ... "ltsp"
22:42
<jammcq>
seems like even 'unknown' would be better than copying the server hostname. although vagrantc's "ltsp" seems like the sanest approach to a default hostname
22:42
<vagrantc>
or "default"
22:42
<jammcq>
copying the server hostname will just confuse people
22:43
<vagrantc>
debootstrap is what populates /etc/hostname ... as it probably doesn't assume it's building chroot environment's for other machines
22:43
<sbalneav>
I don't care about a default if dhcp's not issuing a hostname: it's a garbage default at that point. What I'm trying to solve is that it's not being set EVEN IF the dhcp server's passing it.
22:44
<jammcq>
2 issues then
22:44
<vagrantc>
sbalneav: and i'm telling you, remove /etc/hostname, or zero it out. and it will use the dhcp server's passed values.
22:44
so
22:44
<jammcq>
if dhcp issues a hostname, use it. if not, provide a sane default
22:44
<vagrantc>
like jammcq says
22:45
<sbalneav>
I'll let someone else solve the sane default.
22:45
<vagrantc>
i think the zero'd out /etc/hostname is the best approach
22:46
solved
22:46
if [ '(none)' = "$(cat /proc/sys/kernel/hostname)" ]; then ; echo ltsp > /proc/sys/kernel/hostname ; fi
22:46
YAY!
22:46
i can close two bugs with that!
22:47
<jammcq>
nice
22:47
<vagrantc>
that's assuming that that gets set from dhcp
22:51* vagrantc tries it
22:52
<vagrantc>
gah
22:52
doesn't seem to work
22:54
well, at least, not with a typo
23:03
ok, so the trick works. just need to figure out where to hook it in
23:04
<sbalneav>
are you removing /etc/hostname as well?
23:04
<vagrantc>
sbalneav: thanks for pointing out /proc/sys/kernel/hostname :)
23:04
sbalneav: after i get that patch working smoothly, yes.
23:04
sbalneav: well, i prefer to zero it out, actually.
23:04
makes it easier to bind-mount it if we have to
23:07
<ltsppbot>
"sbalneav" pasted "while read HOSTNAME; do if [ "" (6 lines) at http://pastebot.ltsp.org/299
23:07
<ccjoe>
sbalneav, ltsp-update-sshkeys didn't help, I thought I might have goofed things up, so I tried a reinstall, it does the same thing - workstation boots but I can't login.
23:07
<sbalneav>
I put that at the bottom of ltsp-nbd
23:08
<jammcq>
would /etc/hostname ever contain more than one line?
23:08
<sbalneav>
no
23:08
<jammcq>
would it ever contain ZERO lines ?
23:09
<sbalneav>
it will be "(none)" if not set
23:09
<jammcq>
why the 'while' ?
23:10
<vagrantc>
sbalneav: did you try it ?
23:10
<sbalneav>
yup
23:10
ws250 if hostnames set
23:10
ltsp if not
23:11
<vagrantc>
i was thinking we could put it into nfs-bottom, so it would work with both scripts
23:11
<sbalneav>
jammcq: because, read FOO < /some/file doesn't work for some odd reason.
23:11
vagrantc: right, that's what I don't know, is where to stick it.
23:11
<vagrantc>
my code is only 3 lines :P
23:11
<jammcq>
but ' $(cat /etc/hostname) shouljd do the trick
23:11
<vagrantc>
but for some reason, nfs-bottom wasn't running
23:12
<sbalneav>
mine doesn't call anything external to the shell.
23:12
<jammcq>
although /me is still used to `cat /etc/hostname` syntax
23:12
old school I guess
23:12
<sbalneav>
uhh, jammcq, it was you and I who were doing the read syntax to eliminate external program calls.
23:13
<jammcq>
heh
23:13
<vagrantc>
sbalneav: there's no need to write to /etc/hostname
23:13
<sbalneav>
remember all that optimization we did?
23:13
<jammcq>
seems like there should be a clearer way to suck in a file
23:13
<sbalneav>
vagrantc: then won't the init script just set hostname to what's in /etc/hostname if it's there.
23:14
<vagrantc>
sbalneav: which should be empty
23:14
<sbalneav>
Jeez, I can't seem to do anything right tonight.
23:14
<jammcq>
if 'while' is the most effecient, then might as well use it. it just implies a variable number of records when I look at it
23:14
<sbalneav>
I'll go back to doco.
23:14
<vagrantc>
sbalneav: we can: echo > $ROOT/etc/hostname
23:14
in ltsp-build-client
23:16
<jammcq>
ahhhhh, you are reading /proc/sys/kernel/hostname, not /etc/hostname
23:24jon has joined #ltsp
23:32dr_byson has joined #ltsp
23:33
<dr_byson>
who a install ltsp in archlinux ?
23:33ccjoe has quit IRC
23:38ccjoe has joined #ltsp
23:46
<vagrantc>
ah, executable bit.
23:46
yet, let's drop it in nfs-bottom
23:47
sbalneav: i'll commit this for debian, at least.
23:54
<sbalneav>
vagrantc: Pushed yet?
23:55
<vagrantc>
sbalneav: i'm adding it in my other branches first