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


Channel log from 9 February 2010   (all times are UTC)

00:15alkisg has joined #ltsp
00:28Selveste1 has quit IRC
00:34alkisg has quit IRC
01:12Selveste1 has joined #ltsp
01:18pts has joined #ltsp
01:19pts has joined #ltsp
01:33frederickjh has joined #ltsp
01:46gnunux has joined #ltsp
01:51
<gnunux>
hi
01:51
<Lns>
hi gnunux
01:56cyberorg has quit IRC
01:58cyberorg has joined #ltsp
02:02Kicer86 has joined #ltsp
02:07alkisg has joined #ltsp
02:07cyberorg has quit IRC
02:08cyberorg has joined #ltsp
02:10alkisg has joined #ltsp
02:31cyberorg has quit IRC
02:32cyberorg has joined #ltsp
02:36mikkel has joined #ltsp
02:59jond3rd has joined #ltsp
02:59jond3rd has left #ltsp
03:07cyberorg has quit IRC
03:12Selveste1 has quit IRC
03:15puckettc has joined #ltsp
03:17
<puckettc>
I am using Ubuntu 9.04 and LTSP 5. I have 3 servers serving 75 clients all with local apps running fairly well. The largest of the servers has a fast RAID array serving the home drives to the other two but It also has a problem with Xorg consuming 90% + of CPU any ideas where I should look?
03:19
This server is an HP G6 ??? with 2 quad core processors and 12G of ram
03:19
Is anyone there?
03:20puckettc has quit IRC
03:20puckettc has joined #ltsp
03:21Selveste1 has joined #ltsp
03:21cyberorg has joined #ltsp
03:21
<puckettc>
I am using Ubuntu 9.04 and LTSP 5. I have 3 servers serving 75 clients all with local apps running fairly well. The largest of the servers has a fast RAID array serving the home drives to the other two but It also has a problem with Xorg consuming 90% + of CPU any ideas where I should look?
03:26
<alkisg>
puckettc: just an idea - do you perhaps have both vino-server and italc installed?
03:29
<puckettc>
I do have vino installed but don't even know what is is for. I do not have italc
03:30Selveste1_ has joined #ltsp
03:30Selveste1 has quit IRC
03:40
<puckettc>
Anyone have an idea of the best location to post my question?
03:43
<alkisg>
This is a good location, it just isn't a good time :) Try again in about 6-10 hours...
03:43
bbl
03:43alkisg has quit IRC
03:48rhodan_ has quit IRC
03:48
<frederickjh>
puckettc there is a problem with Xorg the X display manager I am running Karmic and I think that it affects 9.04. I have seen no good permanent solution.
03:49rhodan has joined #ltsp
03:49
<frederickjh>
What I do is log out and back in on the server and normally this drops the CPU activity.
03:53
vino is the remote desktop server. If you do not access your machine this way (VNC) then you might want to uninstall it. I have heard for some this stops the CPU hogging.
03:53yanu has joined #ltsp
03:53yanu has joined #ltsp
03:54
<frederickjh>
Do you have X running on the servers?
03:55Faithful has joined #ltsp
03:58
<puckettc>
I do have x running. I sort of cheated as the desktop install was the easiest.
03:58
<frederickjh>
Try logging out and back in to see if that lowers your CPU usage.
03:59
Normally it works the first time for me.
03:59
<puckettc>
Also I suppose I should be running the server package but I jumped the gun a little with the project. This only occurs on one of the servers so I thought I would ask here.
04:00
<frederickjh>
I have X running on my server but I am not serving as many clients as you.
04:00
<puckettc>
I can ssh into the machine with no one logged on at the terminal and it still is 90 or so percent
04:00
<frederickjh>
So X is not running?
04:01
What does top report when you run it in a terminal?
04:01
<puckettc>
each of my servers handles about 25 clients. I have ssh tunnel disabled in the lts conf
04:01
<frederickjh>
top will tell you you biggest CPU users.
04:01
Are you using encrypted traffic to the clients? (This is the default.)
04:02
<puckettc>
no I have turned that off
04:02
<frederickjh>
You can turn it off in the lts.conf.
04:02
I would first check what top says in a terminal and see what is using the cpu.
04:03
<puckettc>
top has xorg between 50 and 90%
04:04
<frederickjh>
Did you try login out and back in?
04:04
<puckettc>
I'll see what I can find later. I have a class to teach thanks
04:04puckettc has quit IRC
04:25Egyptian[Home] has quit IRC
04:31Selveste1_ has quit IRC
04:39Selveste1_ has joined #ltsp
05:02hersonls has joined #ltsp
05:48Selveste1_ has quit IRC
06:07scottmaccal has joined #ltsp
06:18alkisg has joined #ltsp
06:20pmatulis has joined #ltsp
06:22abeehc has quit IRC
06:29abeehc has joined #ltsp
06:30GodFather has joined #ltsp
06:50etyack has joined #ltsp
06:56Briareos1 has joined #ltsp
06:56
<Briareos1>
does one know how i can find which users VNC-session runs on which port?
06:58zip0 has joined #ltsp
07:05pts has quit IRC
07:06pts_ has joined #ltsp
07:10gnunux has quit IRC
07:23gnunux has joined #ltsp
07:25scottmaccal has quit IRC
07:27
<frederickjh>
netstat will tell you what is using the network and ports
07:28scottmaccal has joined #ltsp
07:31cyberorg has quit IRC
07:33cyberorg has joined #ltsp
07:33
<Briareos1>
frederickjh but netstat shows active connections only, right? i need to see (for example) that the open port 5904 hosts the desktop-session of mr. jones
07:34
<frederickjh>
So Are you looking for open ports?
07:34
<Briareos1>
i can "nmap HOST -p 5900-5999" to find which ports are open
07:34
and where sessions are to be found
07:34
but with nmap i cannot figure out WHO is to be found on that port
07:35
i'd need to know that mjones is located at 5903
07:35
<frederickjh>
So, you are using VNC to monitor thin clients on a LTSP setup?
07:35
Which distro?
07:35shawnp0wers has joined #ltsp
07:37
<frederickjh>
Are you doing this from the server or remotely?
07:39GodFather has quit IRC
07:40
<frederickjh>
Maybe the better question is how does the vnc server assign the ports.
07:40
Does it give them out in the order they are started?
07:40
Or can you set them static in a config file?
07:41
If you are using Ubuntu you can us the ITalc Master Interface to do monitor student easily.
07:41
us=use
07:47hersonls has quit IRC
07:51
<Briareos1>
frederickjh ... it gives them out in order they are started it seems
07:52
i'd have to have another look at italc
07:52
<frederickjh>
Are you using Ubuntu?
07:52
<Briareos1>
but in 8.04 the supplied version is faulty
07:52
yes ubuntu
07:52
<frederickjh>
Ok, I just upgraded to Karmic and ITalc work fine in Karmic.
07:54
I had other problems with Karmic. White noise on the speakers on sound server startup. Had to set switches to turn the mic volumes down to 0 in lts.conf
07:57Selveste1_ has joined #ltsp
08:01
<alkisg>
Briareos1: vnc (vino-server) by default always uses port 5900 and accepts only one user. What guide did you follow to allow multiple vnc connections?
08:01rhodan_ has joined #ltsp
08:02rhodan has quit IRC
08:17zip0 has quit IRC
08:18
<Briareos1>
alkisg I personally didn't enable them - i guess my colleague did it somehow when he installed iTALC. could iTALC enable that?
08:18
<alkisg>
Briareos1: are you talking about vnc connections, italc connections or X connections?
08:18Gadi has joined #ltsp
08:19
<Briareos1>
frederickjh: did you, by chance, also notice that firefox was unusably slow?
08:19
<alkisg>
Briareos1: I think your terminology might be wrong. Can you describe in more detail you do your users connect?
08:19
<Briareos1>
alkisg i am talking about VNC - i guessed iTALC uses vnc connections ....
08:19
<alkisg>
*how
08:19
Yes, iTalc uses encrypted vnc connections
08:19
<frederickjh>
I we use Opera and the problem seems to be more related to java apps in the browser.
08:20
<alkisg>
Briareos1: but that's for monitoring the users. It isn't to be used by the users themselves.
08:20
<Briareos1>
users boot via bootp/tftp (?) and for remote administration i log on the the user's session via "krdc 192.168.1.11:59XX"
08:21
alkisg yes, i am not talking about the users themselves - just for the admins to mirror their desktops for support
08:21
<alkisg>
Briareos1: italc has client autodetection
08:21
You don't need to discover the client addresses manually
08:22
<Briareos1>
can i achieve that autodetection without iTALC?
08:22
<alkisg>
Yes, but that's totally unrelated to vnc
08:22
<Briareos1>
sometimes, when i started krdc directly from a user on the server i could see a list of sessions incl. usernames within krdc
08:22
but since some days that stopped working ....
08:23
<alkisg>
You can't have both italc and vnc installed, they collide with each other (if you don't change the ports manually)
08:23
<Briareos1>
so i guessed there must be some way to do a "discovery" just like krdc sometimes could ...
08:23denethor has joined #ltsp
08:23
<alkisg>
If you do that, you'll have high cpu usage, and frequent hangs...
08:42evilx_ has quit IRC
08:44
<sbalneav>
Morning all
08:45
<alkisg>
s is Scottyyyyyyyyy!!!!!!!!!!one!!!!
08:46
<sbalneav>
Notice to all:
08:46
I figured out how to remove the "only talk if registered" thing last night.
08:46pts_ has quit IRC
08:47
<sbalneav>
so we're back to normal. Don't know why it didn't come off when they upgraded freenode.
08:48evilx has joined #ltsp
08:48
<vbundi>
finally, I can go about my diabolical plan involving not registering my name.
08:54paolino has joined #ltsp
09:02
<paolino>
hello, I just installed ltsp server, but I cannot login from client
09:03
it says incorrect password
09:04
I'm trying to login from console alt-f1, because I cannot read the alt-f7 graphics one
09:06mikkel has quit IRC
09:06litlebuda has joined #ltsp
09:08
<frederickjh>
paolino it will not let you login.
09:08
If you want to you can add a line to your lts.conf and use terminal 2
09:09
<paolino>
it seems to ignore my changes to lts.conf
09:09
<frederickjh>
Where in the one you are using located?
09:09
Are you using Ubuntu?
09:09
<paolino>
yes
09:11
/var/lib/tftpboot/ltsp/i386/lts.conf
09:11
<frederickjh>
That is th correct one.
09:11
SCREEN_02=shell
09:11
SCREEN_07=ldm
09:11
Add that then reboot the client.
09:12
<paolino>
I added X_MODE_0 = 1440x1050 in the default section, but still an unreadeable resolution
09:12
ok, I try that
09:12
<frederickjh>
Then you will have a root shell on terminal 2 no need to log in.
09:13scottmaccal has quit IRC
09:13
<paolino>
mh ctrl-alt-canc is showing a lot of squashfs errors
09:15
ok I have it
09:16XuzhouJim has joined #ltsp
09:22
<paolino>
X_MODE_0 = 1400x1050, is this line what I need to make ldm working ? Or I need to set the card driver also ?
09:24
<frederickjh>
Not sure I have not configured x with lts.conf
09:25GGD has joined #ltsp
09:25
<Briareos1>
sorry, frederickjh & alkisg, that i didn't respond earlier. it's pretty stressy today. I'll read your notes in the evening - thanks.
09:27
<ogra>
paolino, check ~/.xsession-errors
09:27Lns has quit IRC
09:28Lns has joined #ltsp
09:28
<ogra>
paolino, if you get to the login manager all is fine with ltsp, the problem is the desktop or user management on the server
09:28
<paolino>
ogra the login manager is not readable
09:29
<ogra>
oh, i thought you said you get a "wrong passwd" message, sorry, wasnt following closely
09:29
<paolino>
that was on console
09:29
<ogra>
ah
09:30
yeah, thats desired :)
09:30
<paolino>
now I have console
09:30
but ldm is flickering , so wrong resolution I guess
09:31
<ogra>
XHOZRSYNZC and XVERTREFRESH with teh proper values for your monitor should help here
09:32
look in the docs :)
09:32
<alkisg>
!lts.conf
09:32
<ltspbot>
alkisg: "lts.conf" :: http://manpages.ubuntu.com/lts.conf
09:32
<alkisg>
Seach for XRANDR_MODE_0, X_VERTREFRESH and X_HORZSYNC there.
09:33
<ogra>
you shouldnt need XRANDR_MODE_0 though if the latter two are correctly set
09:33
unless i missed code changes :)
09:33
<paolino>
I have ati 1400x1050 lcd
09:34* Gadi reflects on ogra's new position as superbowl commentator
09:34
<ogra>
lol
09:35Selveste1_ has quit IRC
09:35
<paolino>
I've found some modeline parameters googling, can I get the X_VERTREFRESH and X_HORZSYNC, out of them ?
09:35
<ogra>
your monitor handbook surely has them
09:36
and you should be able to find a pdf with google
09:36
<vbundi>
Yesterday Gadi and I were able to get my terminals loading a lot faster by disabling ltspfsd, though my bootchart still doesn't look quite normal http://imagebin.org/84070
09:37
<ogra>
vbundi, well, its readable now :)
09:37
<vbundi>
2 minute boot is acceptable to me, I'm just curious as to why mine looks so much different than the one you sent before
09:37
<ogra>
but you clearly have a CPU issue
09:38
<vbundi>
yeah I disabled the audio hardware and got rid of ltspfsd
09:38
<ogra>
its maxed out all the time
09:38Faithful has quit IRC
09:38
<vbundi>
yeah Gadi thought it might be that the cpu (Transmeta Crusoe) doesn't have much support in the newest kernel
09:38
<Gadi>
well, the jury's still out on all this
09:39
<ogra>
yeah, smells largely like a kernel issue
09:39
<Gadi>
well, it could be a cross-coupled problem
09:39
in other words,
09:39
<vbundi>
so the reason my bootchart looks weird is mostly because it just takes longer to do everything?
09:40
<ogra>
look at the top graph
09:40
the one showing your CPU at work
09:40
<Gadi>
perhaps there are so many udev calls in rapid succession that the ltspfs_entry code cannot handle them all
09:40
<vbundi>
yea I can see it's pretty maxed most of the time
09:40
<Gadi>
and since they run as separate threads, maybe they step over each other
09:40
<ogra>
vbundi, http://people.canonical.com/~ogra/osiris-lucid-20100202-6.png see my laptop compared to that
09:41
<Gadi>
I wonder if we should lock the /var/cache/ltsp_config file that we write to in ltsp_config
09:41
<ogra>
its actually doing stuff while the CPU is maxed out, while yours seems to idle in maxed out CPU states
09:41
<Gadi>
ogra: you prolly have a hyperthreaded CPU
09:41
<ogra>
indeed
09:41
<vbundi>
it's definitely dual core
09:42
<Gadi>
I dont think the transmeta is
09:42
<ogra>
no, it isnt
09:42
<vbundi>
yea transmeta is old
09:42
<ogra>
and i'd like to know the L2 cache
09:42
<vbundi>
1meg I believe
09:42
<Gadi>
so, out the CPU, you max out the CPU
09:42
:)
09:43
<ogra>
well, i dont max out a single core non HT CPU ... at least not to a point where you have 2min boots
09:43
<Gadi>
I think we have to be careful now that we call ltsp_config from ltspfs_entry
09:43
<ogra>
i wonder if we see the same issue as with the SIS CPU in the e1000
09:43
which seemed to be a kernel prob as well and was never solved
09:44Ahmuck has joined #ltsp
09:44
<Gadi>
I think we will see the same issue with all single threaded CPUs
09:44
<Ahmuck>
i'd like to boot from an ltsp server and then "clone" the local hard drive to a folder on the server. possible?
09:45
<Gadi>
which is excellent, because it is resolvable without resorting to differnt packages
09:45
Ahmuck: write a screen script
09:45
or drop to a shell and do it by hand
09:46
<Ahmuck>
a screen script?
09:46
<Gadi>
screen scripts are what gets run at the end of boot
09:46
examples are ldm, rdesktop, shell
09:46
<Ahmuck>
i can drop to a shell and do it by hand. i suppose it's just a matter of mounting the local drive to the server?
09:46
ah, ok
09:46
<Gadi>
all in /opt/ltsp/i386/usr/share/ltsp/screen.d/
09:46
do you want a copy of the files or an image of the drive?
09:48
<Ahmuck>
well, it's a superintdent's laptop. I need to make a backup of the drive in the event repair fails. so a copy of the files and/or an image
09:50
<Gadi>
for files, install rsync in the chroot, boot to a shell, mount the hard drive to say /mnt/sda1 and rsync -avz /mnt/sda1/ user@server:/path/to/folder/
09:50
for an image, you can use dd
09:51
something like: dd if=/dev/sda1|ssh user@server dd of=/path/to/imagefile.img
09:53denethor has quit IRC
09:53Faithful has joined #ltsp
09:56mgariepy has joined #ltsp
09:57mgariepy has joined #ltsp
10:03
<vbundi>
ogra: I've got a terminal with an SiS cpu here
10:03
<ogra>
and ? how many minutes ? :)
10:03
<vbundi>
ogra: it's also got multiple instances of udevd spawned
10:03
one sec.. scpd the tgz to my other system
10:03
<ogra>
we had them down to two ... from five in gutsy
10:05
<vbundi>
4:50
10:05
it's a 200mhz cpu though I think...
10:05
128MB ram
10:05
<ogra>
yeah, the same thing i use as router now
10:06
sad it went back to 4:30
10:06
it was the reason why i switched ubuntu to nbd
10:06
because that shoved off 3min from the boot
10:06
<vbundi>
ah
10:06
<ogra>
and sped up other boots as well :)
10:07
<vbundi>
yea this is one of those little silver/black terminals that can screw into the VESA mount on a monitor
10:07
<ogra>
right
10:07
makes an waesome router if you attach a USB NIC
10:08
<vbundi>
oh yea that'd be kinda cool
10:09
I was going to try and use one for ipcop a while back, for some reason I couldn't get it to work with using usb drive storage
10:09
probably would be better to use CF on this little box
10:11
<ogra>
i have a CF in it
10:11
was hard to get though ... they got expnesive nowadays
10:12XuzhouJim is now known as XuzhouJijm
10:12
<vbundi>
yeah it's cool to do it, but not really practical unless you have stuff layin around
10:13
<XuzhouJijm>
Anyway to get a client to be allowed to administer the server. Servers in a different room. I'm lazy.
10:14paolino has quit IRC
10:14etyack has quit IRC
10:15Ahmuck has quit IRC
10:15Ahmuck has joined #ltsp
10:16
<Ahmuck>
ping
10:17denethor has joined #ltsp
10:22gnunux has quit IRC
10:29
<vbundi>
I managed to get this SIS 128MB terminal booted but the display is quite slow and choppy (screen_02 = rdesktop)
10:30
these terminals did work pretty well before... are they just too slow now for the new kernel?
10:31
<ogra>
define "before" :)
10:31
<vbundi>
ltsp4.. lol
10:31
<ogra>
the slowness showed up around 2.6.20 or so ...
10:31
ltsp4 is left with .18
10:32
<vbundi>
THIS particular ltsp4 setup is 2.4.xx
10:32
<ogra>
well, thats ancient anyway :)
10:33
<vbundi>
yeah well performance is a lot better as far as screen refreshing etc... the main reason I want to upgrade is due to hardware support
10:33
this version won't even work with half the USB mice out there...
10:35
so I guess here's my question: if I don't care about sound, and localdev, or local apps, just want to run rdesktop for these guys so they can use their accounting software, am I better off using LTSP4 or an old version of LTSP5
10:35
^ also, I doubt they want to shell out for new hardware
10:35
so we'll be runnign these Transmeta terminals
10:36
<alkisg>
128 ram is enough, so I guess that's just a driver performance check...
10:37
<ogra>
vbundi, ltsp4 doesnt have security support, no fixes are applied and no development is going on in it at all ... beyond that you would probably be better off with it
10:37Briareos1 has quit IRC
10:37
<ogra>
but its largely a dead project now
10:38
<vbundi>
alksig: most of them are 256MB+ it's just that the Transmeta processors don't seem to be efficient in working with the newer kernel
10:38
<alkisg>
How do you measure efficiency?
10:39
Do you have LDM_DIRECTX=true? is xv enabled? is the driver ok? is your network ok?
10:39
<vbundi>
performace vs expected performance
10:39
network is good
10:39jhutchins has joined #ltsp
10:39dro has joined #ltsp
10:39
<vbundi>
LDM_DIRECTX is not set
10:40
<alkisg>
Did ltsp4 use encryption? I don't think so... so you shouldn't be comparing it with ltsp5 with ldm_directx=false (=the default), because that uses encryption
10:40
<vbundi>
most of these terminals use Radeon chipsets, is there a module to set for them that might increase screen performance?
10:40
<alkisg>
Ssh encryption for video quickly kills your CPU.
10:41
(and lowers performance)
10:42
<vbundi>
alkisg: yeah I understand there are a lot of great things about LTSP5, but if my hardware can't support them, then I have an issue... I will try disabling x over ssh
10:43
<alkisg>
For low spec hardware, disabling encryption will get you a really big performance boost.
10:43
(at the cost of security)
10:43
[Default]
10:44
LDM_DIRECTX=true
10:44
<vbundi>
trying it now
10:44
<alkisg>
that in /var/lib/tftpboot/ltsp/i386/lts.conf or whereever else you have it
10:52
<vbundi>
"LDM_DIRECTX = true" does not seem to affect performance for me
10:52
<johnny>
try without the spaces..
10:52
i also go simple with the Y .. but that shouldn't matter :)
10:52
and yes.. make sure it is under [default]
10:53
<vbundi>
so LDM_DIRECTX=Y
10:53frederickjh has quit IRC
10:54frederickjh has joined #ltsp
10:54
<johnny>
LDM_DIRECTX=true should be fine too tho
10:54
i had probablys with spaces at various times myself..
10:54
<alkisg>
Yup, the problem is in the spaces
10:54
<johnny>
alkisg, what causes the space problem to happen?
10:54
<alkisg>
grammar?
10:54* alkisg hasn't looked at this..
10:55
<johnny>
gadi keeps telling me there is no problem with spaces
10:55
but i'm pretty sure i've seen ti..
10:55
<alkisg>
Ah. I don't know, I just heard it here
10:55
<vbundi>
that's kinda weird... I've got spaces in my other lines and they work
10:55
that's kinda weird... I've got spaces in my other lines and they work
10:55
oops uparrow + enter meant for terminal
10:56
<alkisg>
vbundi: distro/version?
10:56
<vbundi>
running LTSP? ubuntu 9.10
10:56
x64
10:57
<alkisg>
!learn localxterm as while sitting on a thin client, open a gnome terminal. In that, run: ltsp-localapps xterm. An xterm will open. That xterm runs locally, so any commands you enter there are executed directly on the client.
10:57
<ltspbot>
alkisg: The operation succeeded.
10:57
<alkisg>
vbundi: in a localxterm run: getltscfg -a
10:57
!localxterm
10:57
<ltspbot>
alkisg: "localxterm" :: while sitting on a thin client, open a gnome terminal. In that, run: ltsp-localapps xterm. An xterm will open. That xterm runs locally, so any commands you enter there are executed directly on the client.
11:11
<vbundi>
alkisg: hmm my ldm screen seems to be broken
11:11
<alkisg>
How so?
11:11
<vbundi>
won't let me type in it or anything
11:11
<alkisg>
try: sudo ltsp-update-kernels, and reboot the client
11:11
<vbundi>
I haven't tried gnome since I recreated this chroot,
11:11
k
11:12grantk has joined #ltsp
11:12Lns has quit IRC
11:14alexqwesa has quit IRC
11:14
<grantk>
does anyone know if printer declaration has changed since ltsp4.2 for clients?
11:14
I took a look at the lts.conf man and it does not look like it has.
11:14alexqwesa has joined #ltsp
11:14
<ogra>
!docs
11:14
<ltspbot>
ogra: "docs" :: For the most current documentation, see https://sourceforge.net/apps/mediawiki/ltsp/index.php?title=Ltsp_LtspDocumentationUpstream
11:15
<ogra>
grantk, ^^^
11:22ltspbot has joined #ltsp
11:23GGD has quit IRC
11:26XuzhouJijm has quit IRC
11:27sbalneav has joined #ltsp
11:28Briareos1 has joined #ltsp
11:28
<Briareos1>
re
11:29cliebow has joined #ltsp
11:29
<Briareos1>
frederickjh: did you also try firefox 3.5 or 3.6 to see if they work okay within ltsp?
11:29
<cliebow>
a
11:29
<frederickjh>
Can't say for sure, but maybe I could test later.
11:30* frederickjh goes to eat
11:31
<Briareos1>
frederickjh 3.0.x is unusable, but i was hoping that 3.5 - which was advertised as less resource-intensive - would work
11:31Lns has joined #ltsp
11:32
<_UsUrPeR_>
hey all. I am trying to allow users a passwordless login for multiple login names on a single ltsp server. Has anybody attempted this?
11:33
<alkisg>
Briareos1: firefox works fine for me. Is your setup OK?
11:33jhutchins_lt has joined #ltsp
11:33
<alkisg>
_UsUrPeR_: if you set a different LDM_USERNAME and LDM_PASSWORD for each host, it should work fine...
11:34
<_UsUrPeR_>
alkisg: was not looking to bypass LDM, I just want to allow users to login without a password (i.e. click their name and go)
11:35
<alkisg>
Urm, first of all, ldm doesn't offer a users list, does it? :D
11:36
<_UsUrPeR_>
alkisg: on the server it does....
11:36
not the clients though.
11:36
<alkisg>
That's gdm on the server
11:36
Not ldm
11:36
<_UsUrPeR_>
ahh
11:37GodFather has joined #ltsp
11:37
<_UsUrPeR_>
onoz
11:37
<alkisg>
It shouldn't be too difficult to provide a users list, but someone would have to code it... :)
11:38
<johnny>
well.. it'd be nice if we could reuse gdm :(
11:38
or reuse code from gdm...
11:39
<alkisg>
Or let the whole login screen run on the server :)
11:39
(== gdm itself)
11:40paolino has joined #ltsp
11:40
<johnny>
alkisg, hmm.. that would be neat.. would still need to plug certain things
11:40
guess if we had that dbus mapping from client to server it might work..
11:40
<alkisg>
Yeah, we'd need some pam* modules/hooks etc
11:41
<johnny>
and a way to handle the ltspfs mounts
11:41
<alkisg>
It'd be even better if those modules were a normal package, one you could install to any pc and make it an ltsp client
11:43pippo1 has joined #ltsp
11:45
<Briareos1>
alkisg: about ffox - how many users are using firefox concurrently?
11:45
<alkisg>
12
11:50
<cliebow>
brb
11:50cliebow has quit IRC
11:51
<pippo1>
hi , how I decide which applications are run locally when I enable LOCAL_APPS ?
11:52alexqwesa has quit IRC
11:53
<pippo1>
found: LOCAL_APPS_WHITELIST
11:53alexqwesa has joined #ltsp
11:53
<alkisg>
pippo1: by default, anything that has a .desktop file, runs locally
11:54
<Gadi>
alkisg: only if you make LOCAL_APPS_MENU=True
11:54
<alkisg>
If for some reason you want to install an application to the chroot but *not* have it run locally, *then* you use the LOCAL_APPS_WHITELIST
11:54
<Briareos1>
huh? does that mean by default pretty much of the performance relies on the client?
11:54
<alkisg>
Ooops, and what Gadi said :)
11:54
<Gadi>
pippo1: LOCAL_APPS_MENU=True overrides menu entries to launch locally
11:54
pippo1: LOCAL_APPS_MENU_ITEMS=<list> overrides only those in the list
11:55
<pippo1>
alkisg: what is a .desktop ?
11:55
<alkisg>
ls /usr/share/applications
11:55
<Gadi>
pippo1: LOCAL_APPS_WHITELIST=<list> only allows certain commands to run locally
11:55
<alkisg>
The menu files are .desktop files
11:58
<pippo1>
ok that dirs contains almost any appication
11:58
<grantk>
other than the lts.conf for enabling printing to a terminal with ltsp 5 on ubuntu is there anything else to do. I checked the LTSPManual.pdf and my declarations in my lts.conf are correct, did not change from 4.2 actually. I looked at the ubuntu community documentation page and did not find anything there. One thing I know for sure is that I can not telnet to the clients that I have defined a printer for.
11:58
<alkisg>
Argh... nbd doesn't work for me with vbox clients in Lucid :(
11:59
<grantk>
http://pastie.org/816678
11:59
<pippo1>
Gadi: that means if I set LOCAL_APPS_MENU=True thy are all run locally ?
12:00
<dro>
alkisg: I think when I had issues with vbox in the past, I used nat for the nic and then nbd worked
12:00
<grantk>
alksig: do you have to use a boot image for those vbox clients or are you able to get them to pxe boot?
12:00
<Gadi>
pippo1: anything in /opt/ltsp/i386/usr/share/applications
12:00
grantk: remove the inline comments next to the MAC addys
12:00
<alkisg>
dro: hmmm thanks I'll give that a try :)
12:00
<grantk>
gadi: ok
12:01
<alkisg>
I'll also try to start nbd as a daemon, not from inetd...
12:01
<dro>
alkisg: i'm actually building a 9.10 ltsp server now in vbox
12:01
<grantk>
ahh
12:01
<dro>
alkisg: when it's done i'll play around and see what the trick was if nat wasn't it
12:02
<grantk>
I have my new servers running on xenserver with nbd swap
12:02
<pippo1>
Gadi ok, I want to run only mplayer.desktop. LOCAL_APPS_MENU_ITEMS= mplayer.desktop ?
12:03
<Gadi>
pippo1: do you have mplayer installed in the chroot?
12:03
<_UsUrPeR_>
alkisg: I'm attempting to use LDM_USERNAME = "testuser1" and LDM_PASSWORD = "test"... doesn't seem to be working properly.
12:03
do I have the nomenclature proper?
12:03
<alkisg>
You also need LDM_AUTOLOGIN=true
12:03
<_UsUrPeR_>
ahh~~~
12:03
<pippo1>
Gadi, I suppos not
12:04
<Gadi>
pippo1: step 1: chroot /opt/ltsp/i386 apt-get install mplayer
12:04
<_UsUrPeR_>
alkisg: do I need that under [DEFAULT] or [<MAC ADDRESS>]?
12:04
<alkisg>
_UsUrPeR_: or LDM_GUESTLOGIN=true if you want to get a guest button for those credentials
12:04
<Gadi>
pippo1: step 2: sudo ltsp-update-image
12:04
pippo1: step 3: set LOCAL_APPS_MENU=True
12:04
<alkisg>
_UsUrPeR_: the LDM_AUTOLOGIN can go under [Default], the others under the mac addres
12:05
<Gadi>
pippo1: step 4: (optional) if all you want to run locally is mplayer, regardless of what other apps you installed in the chroot, you can limit it with LOCAL_APPS_MENU_LIST="mplayer"
12:05
(the names you use in the list are the names of the .desktop files (without the .desktop extension)
12:06
<alkisg>
stgraber: would nbd-proxy have problems with vbox?
12:06* pippo1 cut and paste instructions :)
12:06
<_UsUrPeR_>
is there anything I can do with lts.conf to auto start firefox on user login?
12:07
<alkisg>
To autostart programs, you can just put them in /etc/xdg/autostart...
12:07
<_UsUrPeR_>
hmm k
12:08
<alkisg>
stgraber: is there any switch/kernel parameter that I can use to try without nbd-proxy?
12:10
<stgraber>
alkisg: easiest is to patch the initrd manually, edit /usr/share/initramfs-tools/scripts/ltsp_nbd, comment the nbd-proxy line and update the nbd-client call to use the server IP address and not localhost
12:10
<alkisg>
stgraber: thanks!
12:10
<grantk>
Gadi: No luck with those inline comments removed, I still get rejected when I try netstat or telnet to those clients.
12:10
<Gadi>
grantk: you do telnet <ip> 9100
12:10
<grantk>
yep, and nc <ip> 9100
12:11
<Gadi>
note: *not* <ip>:9100
12:11
<grantk>
telnet 192.168.1.1 9100
12:11
Trying 192.168.1.1...
12:11
telnet: Unable to connect to remote host: Connection refused
12:11
<Gadi>
what distro?
12:11
<grantk>
ubuntu 9.04
12:12etyack has joined #ltsp
12:13
<Gadi>
grantk: firewall?
12:13
<alkisg>
stgraber: btw, for some strange reason, update-initramfs -u *executes* the scripts! E.g. " /tmp/mkinitramfs_Ygsd8R/scripts/nfs-bottom/ltsp: 5: cannot open /proc/sys/kernel/hostname: No such file"
12:13
<grantk>
no
12:14
<Gadi>
grantk: drop to a shell and make sure jetpipe is running
12:14
<stgraber>
alkisg: is that with my PPA ?
12:15
alkisg: and do you have /proc mounted in the chroot ?
12:15
<alkisg>
stgraber: with your ppa packages copied in my ppa, yes...
12:15
stgraber: no
12:15
Erm sorry, I guess on this chroot I didn't use your ppa: ltsp-client 5.1.99-0ubuntu1
12:16* alkisg looks again..
12:16
<grantk>
gadi: no jetpipe
12:16
looks like there is no package so I will try and find the source
12:16
<Gadi>
should be in /usr/bin/ or /usr/sbin
12:16
it is our own script
12:16
one sec
12:17
<stgraber>
alkisg: I fixed that issue upstream a few days ago so it's probably not in Lucid
12:17
<alkisg>
stgraber: just checked with another chroot with your ppa in it, it's OK. Thanks!
12:18
<stgraber>
np
12:18pippo1 has quit IRC
12:18
<Gadi>
grantk: can you run: jetpipe /dev/lp0 9100
12:19
<alkisg>
stgraber: yeah, nbd-proxy has problems with my vbox. Removing it made my vbox work fine.
12:19
<Gadi>
at the shell?
12:19
<grantk>
nope, I just searched for jetpipe and I am not locating it, checked /usr/bin and /usr/sbin
12:19
<Gadi>
in the chroot?
12:19
<grantk>
looks like it did not get installed
12:19
yes
12:19
<Gadi>
it isnt a separate package
12:19
<grantk>
hm strange
12:20
<Gadi>
can you update the chroot?
12:20
ltsp-client should pull it in
12:20
and make sure it also pulls in python-serial
12:20
<grantk>
just an aptitude update?
12:20
in chroot
12:20
<vbundi>
alkisg: hey earlier you wanted some output of getltscfg -a, here it is: http://pastebin.org/89005
12:21
<Gadi>
apt-get update; apt-get upgrade
12:21
<grantk>
0 updated there
12:21japerry has joined #ltsp
12:21
<alkisg>
vbundi: ldm_directx should be in effect - you don't see any major perfomance boost with that?
12:22
<grantk>
there is no python-serial either
12:23
<Gadi>
and no /usr/sbin/jetpipe?
12:23
<grantk>
no
12:23
I was able to install python-serial
12:23pippo1 has joined #ltsp
12:23lns_ has joined #ltsp
12:24
<Gadi>
well, update the image and see where you stand
12:24
<grantk>
oik
12:24
ok
12:24
<pippo1>
Gadi, it seems ok, mplayer runs locally , but cannot open files
12:24
<Gadi>
pippo1: local apps will see the user's home directory
12:24
<vbundi>
alkisg: not particularly, I'm wondering though if the proper module is being loaded for the display, I had a problem like that in LTSP4 where I had to specify the module or else scrolling was very choppy
12:25
<Gadi>
pippo1: but the rest of the system is the chroot
12:25
<vbundi>
alkisg: these are radeon chipsets for video
12:25
<alkisg>
vbundi: you should even have a faster boot process...
12:25
<Gadi>
pippo1: so, it should see files in the homedir
12:25
<alkisg>
vbundi: sorry, I've no idea about radeon video modules
12:27
<vbundi>
is there a way to ask the xserver which module it's using?
12:27
<pippo1>
Gadi: can I symlink another from home to access it ?
12:28
<Gadi>
pippo1: local apps see home through sshfs
12:28
by default, sshfs (like ssh) will follow symlinks
12:28
so, if you symlink /home/user/tmp -> /tmp
12:28
<alkisg>
vbundi: I think `lspci -vv -nn` in the localxterm can tell you that
12:28
<alincoln>
vbundi: search /var/log/Xlog.7.log for 'LoadModule'
12:28
<Gadi>
you will see the server's /tmp
12:28
<alincoln>
or maybe that.
12:28
<vbundi>
ok
12:30
<Gadi>
pippo1: at one point someone wrote some additional code to mount more than just the homedir for the client to see, but I am not sure if that made it upstream
12:30
<alincoln>
Gadi: it is upstream, and i'm playing with it now
12:30
not working for me yet
12:31
pippo1: ^^ it's LOCAL_APPS_EXTRAMOUNTS
12:31
<stgraber>
alkisg: weird ...
12:31
<alkisg>
stgraber: I tried with the other chroot kernels, with your ppa in my sources, the same problem...
12:31
When I remove nbd-proxy, it works fine
12:31* alkisg tries with a 32bit client now...
12:31
<stgraber>
alkisg: do you see it connecting to the server ?
12:32* stgraber admits he never tested 64bit with LTSP
12:32
<vbundi>
what's the motivation behind running a 64-bit client?
12:32
4GB ram for local apps? ;)
12:33pippo1 has left #ltsp
12:33
<vbundi>
err 4+
12:33* alkisg used 64bit virtual clients because he had an amd64 server, and for some reason 32bit virtual clients didn't boot! :)
12:33
<alkisg>
But now I'm with i386 on the server again...
12:34
stgraber: give me 15 minutes to do something else that I have to do, then test properly, and report back.
12:40pippo1 has joined #ltsp
12:42
<frederickjh>
Briareos1: I just tried firefox 3.5 vs Opera 10.10 to see if they work okay within ltsp.
12:42
<Briareos1>
yes?
12:42
<frederickjh>
I have a P4 server and only one client was running.
12:42
Plus a gui and stuff on the server.
12:43
Firefox bounced around more from the up 30
12:43
30% to low 90%
12:43
While Opera was more steady between 40% and 60%.
12:43
This was while viewing the same webpages.
12:44
Intel P4 3.2GHz
12:44
I hope that helps!
12:45
<Briareos1>
what are the percentages? cpu usage? mem?
12:45
<frederickjh>
I have been using Opera for years now and find it much lighter.
12:45
I used the GUI system monitor and that was the CPU usage.
12:46dro has quit IRC
12:47
<frederickjh>
With the client down and just what was running before I am running at 40 some % right now.
12:47
<Briareos1>
hmm. our ltsp is on a 2 x quad core ibm machine with 8 GB ram ... but still with about 10 users logged on, firefox hangs the whole users' session at times !
12:47dro has joined #ltsp
12:47
<grantk>
Gadi and Ogra: Thank you.
12:47
<Gadi>
grantk: it works?
12:47
<alkisg>
stgraber: I tried with an i386 virtual client, the same problem: I do get a connection, but then it hangs. Last messages: 'Hangup \n Negotiation: ..size = 2085456KB \n bs=1024, sz=2085456 \n [ 3.5678237] nbd0: unknown partition table \n [ 3.600873 ] aufs test_add:243:exe[226]: uid/gid/perm /rofs 0/0/0755 , 0/0/01777"
12:47
<grantk>
probably if I did not forget to chmod jetpipe
12:47
<frederickjh>
I would try to get the users to try Opera. It is much lighter and faster that Firefox.
12:48
<grantk>
I found this: https://bugs.launchpad.net/ubuntu/+source/ltsp/+bug/224259
12:48
it has a copy of jetpipe ogra posted back in 2008
12:48
<frederickjh>
Briareos1: have you considered ltsp-cluster? It is out now for Ubuntu.
12:48
<grantk>
I put it in /usr/sbin and /usr/bin and then rebuilt the image and now I can telnet to those hosts.
12:48
sorry clients
12:49
<dro>
alkisg: still having nbd issues?
12:49etyack has left #ltsp
12:49
<alkisg>
dro: yeah, vbox doesn't cooperate with a new nbd-proxy thing
12:49
<grantk>
so I need to change permissions on it and will hopefully be in business
12:49denethor has quit IRC
12:49
<Briareos1>
frederickjh: you mean more than one server for 10 clients?
12:49
<dro>
alkisg: my client is almost done building, i'll tinker with it after that
12:50
alkisg: which version of lucid did you download, the daily build?
12:50
<alkisg>
dro: Karmic? That new nbd-proxy thing is Lucid - specific, so we won't be having the same problem...
12:50
<Briareos1>
could that, at least also be used as a backup in case one server goes down, frederickjh? in that case i'd consider it ...
12:51
<frederickjh>
Yes, however if what they are mostly doing is browsing the web then maybe changing browser is a better idea.
12:51
<alkisg>
dro: I have stgraber's ppa in my sources - they're not in lucid yet...
12:51
<dro>
alkisg: i know but i know there were nbd issues with karmic on vbox too, and i found a way to fix it without having to use nfs
12:51
<frederickjh>
Have you turned of encryption of traffic between the server and clients in the lts.conf?
12:51
<alkisg>
dro: I didn't have any nbd problems with vbox and karmic...
12:51
<frederickjh>
Than will save some CPU cycles.
12:51
<dro>
alkisg: maybe i just had bad luck lol
12:51
<alkisg>
:)
12:52
<dro>
alkisg: of course i've slept a few times since then maybe it was intrepid
12:53puck1 has joined #ltsp
12:53
<frederickjh>
LDM_DIRECTX=True in lts.conf will turn encryption off Briareos1
12:55
<Briareos1>
frederickjh ok i'll try that too. the firefox-problem could just as likely be bandwidth-related - don't you think?
12:56
<frederickjh>
For more info on ltsp-cluster see their new site https://www.ltsp-cluster.org/home
12:56
Not really. I was surfing the same sites.
12:58
<puck1>
I need help with a fairly complex LTSP setup. I have 3 servers with about 75 clients. I have one main server that has the home drive for all users. This is shared via NFS between the servers on a server only network segment. The clients are slow in responding when most are running but I can't tell if the problem is a lack of processor power or write /read time to the home drive. The main server has 2 quad core 2.2Gh processors and 12Gig of memory. The othe
12:58
<frederickjh>
Try an experiment ask your user to all use Opera for 10 minutes and see what happens to the load compared to firefox.
12:59
<puck1>
All of the servers are connected via Gig ethernet to decent 3com switches
13:00
<vbundi>
be back in an hour
13:00
<sbalneav>
puck1: Thing to check would be most processes, when they're slow, if their status is "D" in ps
13:00
is so, then they're blocked on io
13:01mikkel has joined #ltsp
13:01
<puck1>
What is the D status in the process list
13:01
<sbalneav>
If you've got home + NFS on a server that's ALSO serving clients, you have a potential for a real problem
13:01
Process status
13:01
<puck1>
should I move the home to a fast box by itself
13:02
<sbalneav>
I'd say that'd be at least a start.
13:02
As well, running X and NFS over the same ethernet segment may cause you problems.
13:02
<puck1>
Is this a viable plan to server 100 or so clients. Ubuntu + LTSP
13:02
<sbalneav>
X is chatty, and sends lots of small packets
13:02
NFS wants big, chunky packets.
13:03
<puck1>
I am not running anything except NFS over the backend
13:03
<sbalneav>
puck1: I've got about 100 thin clients here at my location.
13:03
exactly the setup you describe
13:03
with 3 LTSP servers
13:03
<puck1>
sbalneav: How many servers?
13:03
<sbalneav>
1 NFS server
13:04
Each LTSP server's a quad xeon, 8 gigs ram
13:04
NFS server's a quad with 4 gigs
13:04
all servers are networked together with a backbone segment, over which just NFS and internet access goes.
13:04
<puck1>
Does your home box do anything else?
13:04
<sbalneav>
Nope
13:04
just NFS
13:05
6 disks
13:05
raid 10
13:05
it's running flat out.
13:05
<puck1>
How are the ltsp servers connected?
13:05
<sbalneav>
Gig backbone
13:06
all terminals hang off their own interface to each server
13:06
1 sever has 40
13:06
2 servers have 30
13:06
<puck1>
How many connections do you have to the NAS and backbone?
13:06
<sbalneav>
Don't have a nas
13:06
NFS server has it's own built in disks
13:06
<puck1>
Sorry wrong terminology
13:06
<sbalneav>
each ltsp server has 2 network interfaces
13:06
<puck1>
Same here
13:06
<sbalneav>
1 for backbone (nfs, internet)
13:07
1 for terminals
13:07
<puck1>
one for home one for everything else
13:07
I have all terminals in the building on the same switched network so I can flip them between servers as needed.
13:08
<sbalneav>
I'd segment them
13:09
<puck1>
Do you use any local apps?
13:09
<sbalneav>
None
13:09
All off the server.
13:09
<puck1>
which flavor of linux
13:09
<sbalneav>
Firefox, Openoffice, Thunderbird for mail
13:09
couple of wine apps
13:09
Ubuntu Hardy, ATM
13:10
<Briareos1>
frederickjh: i mean bandwidth thinclient-to-ltspserver
13:10
<puck1>
Ubuntu 9.04 with LTSP are you using the server build
13:11pippo1 has quit IRC
13:11
<puck1>
Do you have any problems with speed while running large apps like GIMP?
13:11
<frederickjh>
Not sure I had one client running on a 10Mpbs connection. But maybe you are referring to your setup.
13:12
What do you have 1 Gbps to the switch and the 100Mbps to the clients Briareos1?
13:12
<sbalneav>
puck1: We don't run a lot of the gimp, but when I fire it up, I don't notice anything really bad
13:13
Panning around in a large image is obviously going to be a bit slower on a thin client than a full workstation.
13:13
<Briareos1>
frederickjh: ltspserver - gigabitswitch - thinclients (every client on it's own jack directly on the gbitswitch)
13:13
<sbalneav>
Thin clients are not meant to be comlete substitutes for full fledged workstations.
13:13
<puck1>
OK I will try to move the NFS to a server by itself
13:13
<frederickjh>
So from the switch to the clients you have giabit ethernet?
13:14
<puck1>
I am just having problems saving files not viewing them so I think the NFS problem is where to start
13:15
<frederickjh>
Briareos1 have you tried monitoring the network connection to the switch to see if it is getting/staying full?
13:15
<puck1>
We do a lot of GIMP and other editing tools here in our school so I have to do something to fix this
13:15
<sbalneav>
yeah, if it's file access, your NFS server's prolly saturated
13:15
<Briareos1>
frederickjh: hmm i wouldn't really know how to monitor that ... (?)
13:15
<sbalneav>
make sure to run more NFSd's than the default (8)
13:15
I run 64 of them
13:15
<puck1>
Any sugestions on how to ser one up well
13:16
<sbalneav>
There's some sysctl stuff you can do to tweak, if you need to.
13:16
hold on...
13:17
<frederickjh>
Briareos1 try iptraf and general interface satistics.
13:17
It is a great tool for network monitoring.
13:17
<sbalneav>
puck1: Add RPCNFSDCOUNT=48
13:18
<puck1>
My two additional servers connect to the NFS box by direct crossover cables gig ethernet to gig ethernet
13:18
Does this sound like a problem?
13:18
<frederickjh>
If it is full maybe you should consider adding another gibabit interface from the server to the switch and bonding them together. This would double your capacity for each client.
13:18
<sbalneav>
to your /etc/default/nfs-kernel-server
13:19
puck1: should be fine
13:19
<puck1>
sbalneav: then just restart NFS??
13:20
<sbalneav>
Don't do it on the one you've got now
13:20
do that for when you set up the new server.
13:20
<puck1>
what would happen if I tried on this?
13:20jammcq has joined #ltsp
13:20
<jammcq>
hey friends
13:21
<sbalneav>
!j
13:21
<ltspbot>
sbalneav: "j" :: jammcq!!!!!!!!!!!!!!!!!!!!!!!
13:21
<jammcq>
Scotty !!!!!!!!!!!!!!!!!!!!!!!!!!!
13:21
so we're getting a little bit of snow around here. there's probably 1/2 inch so far, with about 2" total for today and another 5" over night.
13:21
people are just freaking out
13:22
several suggestions to cancel the user group meeting tonight because of it
13:22
i'm thinking.... people... get a grip
13:22
<alincoln>
jammcq: hah, you like the remote meeting suggestion on the mailing list?
13:23
<jammcq>
I think there's some value to showing up face-to-face
13:23
<alincoln>
you bet.
13:23
<puck1>
sbalneav: I bought a big HP server to try to do both for about 20 users and it works OK but when serving the additional servers it is just not working. Thanks for your help
13:23
<jammcq>
a remote meeting will almost completely eliminate audience participation
13:23
<alincoln>
i thought it was funny how the snow spawned this sudden interest in remote meetings.
13:24
<sbalneav>
puck1: NP
13:25
<Briareos1>
frederickjh: great hint! i just installed it and have it running right now
13:25
<puck1>
3
13:25
.
13:25
<Briareos1>
frederickjh: though there are no users at that time of day :)
13:25
<frederickjh>
Briareos1 iptraf or the gigabit interface?
13:26
<Briareos1>
frederickjh: iptraf
13:26
<Gadi>
alkisg: ping
13:26
<alkisg>
Hey Gadi
13:26
<frederickjh>
So you have no user now?
13:26
<Briareos1>
frederickjh: no
13:26
<Gadi>
hey - are you aware of any kernel command args to set mtu in the initramfs?
13:26
<frederickjh>
Are you in europe?
13:26
<Briareos1>
frederickjh: yes :)
13:26
<frederickjh>
Like me?
13:27
Well there always is tommorrow.
13:27
<Briareos1>
frederickjh: server has 2 interfaces - one going out to the "world" (not directly ofc) and one to the gbit-switch with only ltsp-users attached to it
13:27
<alkisg>
Gadi, I'm not even sure what mtu exactly does :D
13:27
(ok I know a little)
13:27
Gadi: there's a kernels-parameters.txt, did you look at that?
13:28
http://www.kernel.org/doc/Documentation/kernel-parameters.txt
13:28
<frederickjh>
If the gbit gets full you can add a second one in parallel and bond the two together. bingo 2gigabit interface.
13:28
<Briareos1>
frederickjh: after what happened today i am pretty afraid of tomorrow :] having some issues with our ERP-system ... and i'm not sure if it's related to the samba PDC ... and also the firefox-thing causing clients to crash ... and so forth ...
13:28
<Gadi>
yeah, no joy - I think it is something I will need to code
13:29
<Briareos1>
frederickjh: that sounds really good - and with iptraf i am finally able to find out where the cause actually is to be found
13:29
<Gadi>
MTU = the maximum size of a packet your machine will send out its interface
13:29
<frederickjh>
Business or school?
13:29
<Briareos1>
frederickjh: biz
13:29
<Gadi>
default is 1500 on linux/ethernet
13:29
<frederickjh>
Yes that is a bit scary.
13:30
<grantk>
Briareos - I have not seen everything that you have been saying but I had firefox causing issues when I was not using nbd_swap, so I was always running into oom errors and the xsession would get killed.
13:30
<Briareos1>
frederickjh: i just started iptraf in a "screen" session, so i already have some data tomorrow i hope
13:30
grantk: nbd_swap?
13:31
<grantk>
yep, most of my clients are pretty old and only have 64 -128 mb ram so I stated using nbdswap
13:31
just a few lines in the lts.conf and elsewhere.
13:31
briareos1: do you happen to be using ubuntu:
13:32
<Briareos1>
grantk: yes
13:32
grantk: we especially are having issues with one client thats older/weaker than the others
13:32
grantk: so that may be an issue
13:32
<grantk>
yep, that was my problem. I got tipped onto the memory issue by replacing one with a newer client that has 1gb mem
13:32
I have my instructions around here somewhere, hold up
13:33
<Briareos1>
grantk: why is there so much memory usage on the client-side anyways?
13:33
<alkisg>
Gadi: well a simple ifconfig or ip call in an rc script should do it, right?
13:33
<grantk>
I am not too sure
13:33
I just noticed it was usually with people browsing the web in firefox
13:34
I also tried to switch as many users as I could to opera
13:34
it has smaller memory footprint, and limit the number of firefox addons people use
13:34
mozilla thunderbird was a problem as well
13:34
<Briareos1>
grantk: i can pretty much delegate which browser is to be used in our environment, but i'd risk new issues with websites not displaying correctly, flash, java, ... (?)
13:35
<frederickjh>
Opera 10.10 is pretty good compared to the old days.
13:35
<grantk>
yeah, that would hold you back, I run 64bit here so I rely on that to keep me from worrying about flash right now
13:35
<frederickjh>
I started using Opera back in 7.0 days.
13:35
<grantk>
but on my laptop I use here I know that flash works just as good in chrome as it does in firefox
13:35
and chrome uses less memory
13:36
<Briareos1>
grantk: chrome's not an option due to privacy reasons :)
13:36
<alkisg>
Woah, gpxe 1.0 is out and has wireless support!
13:36
<Gadi>
alkisg: I am trying to decide whether it needs to be done in initramfs (ie if mounting nbd would fail)
13:37bobby_C has joined #ltsp
13:37
<Gadi>
alkisg: I am inclined to think no
13:37
<alkisg>
Gadi: why do you need to change it? Speed reasons?
13:37
<Gadi>
alkisg: and it is safe to do it in rootfs-space
13:37
alkisg: it avoids fragmentation
13:37
<grantk>
heh, chrome crashed on me, hold on i found those instructions!
13:37
<Gadi>
alkisg: think of it this way
13:38
thin client sends packet size 1500 through router that only accepts MTU 1492
13:38
so packet need to be broken into two
13:38
and then reassembled
13:38
<grantk>
http://pastie.org/816851
13:39
Briasreos1: That is missing the file size declaration, hold on there
13:39
<johnny>
hmm..
13:39
<alkisg>
Gadi: OK, so in the end, that's a speed reason. So you could probably do it any time, even after the clients boots... so whereever's easier, e.g. in an ldm rc script?
13:39
<Gadi>
MTU is largely irrelelvent with the speed of modern network equipment, tho it can still impact networks with older equipment or gobs of traffic
13:39
<grantk>
https://help.ubuntu.com/community/UbuntuLTSP/EnableNBDSWAP
13:39
<johnny>
looking for 512mb for my clients.. what's the deal with ecc mem in non ecc using machines?
13:39
<Gadi>
alkisg: right
13:39
I am thinking the same
13:39
<grantk>
there is a /et/ltsp/nbdswapd.conf that needs created
13:40puck1 has left #ltsp
13:40
<grantk>
Briareos1: here is an example of the swap files once created, they go to /tmp on the server
13:40
-rw------- 1 nobody nogroup 268435456 2010-02-08 19:38 tmp.AujJRBYVsb
13:41
those instructions were for ubuntu 8.04 but I used them with 9.04 and had no problems.
13:41
<Briareos1>
grantk thank you ... will the client's ram still be used i guess?
13:42
<grantk>
yes
13:42
<Briareos1>
grantk: if, on the other hand, i update all clients to about 1 GB RAM - would that be the better solution (performance-wise)?
13:42
<grantk>
I am not sure, I have not watched my clients with that much memory(only 3 of them!)
13:43
I imagine so since memory usually is better than swap
13:43
<Briareos1>
grantk: ok. ... and nbd_swap will use the server's RAM primarly (if available) or will it go directly to a file?
13:43
<grantk>
from my understanding directly to the file in /tmp
13:44GodFather has quit IRC
13:45
<Gadi>
nbd_swap is swap space for processes running on the client
13:45
it supplements client memory not server memory
13:45
for things like Xorg (when it loads lots of bitmap images into client memory)
13:47
<dro>
alkisg: you're right I didn't have any issue with karmic on vbox, of course I set the network to host only adapter
13:49
<Briareos1>
Gadi ty
13:52rhodan_ is now known as rhodan
13:52
<johnny>
alkisg, do you know if it's ok to put ecc memory in a non ecc mahine?
13:53
obviously i don't expect it to work with ecc :)
13:53
i read something that says it should on most machines
13:53
just wondering if anybody had any personal experience
13:54
<alkisg>
johnny: I think I tried that but back with SDRAM, not with DDR...
13:55
It worked fine, if my memory serves me.
13:57
<johnny>
sdram yes
13:57
ok
14:07vagrantc has joined #ltsp
14:13etyack has joined #ltsp
14:22bobby_C has quit IRC
14:25bobby_C has joined #ltsp
14:31Selveste1_ has joined #ltsp
14:31bobby_C has quit IRC
14:32bobby_C has joined #ltsp
14:42Kicer86 has quit IRC
14:44GGD_ is now known as GGD
14:57ltspbot has joined #ltsp
15:00
<vagrantc>
stgraber: plans on tagging/uploading ltsp/ltsp/ltspfs for lucid still, or are you hitting freeze?
15:08shogunx has quit IRC
15:10
<alkisg>
vagrantc, so is also apache going to change to using /srv?
15:10
<vagrantc>
alkisg: haven't seen it happen yet.
15:10
<alkisg>
stgraber started a topic on the ubuntu devel ml for this: https://lists.ubuntu.com/archives/ubuntu-devel/2010-February/030197.html
15:12shogunx has joined #ltsp
15:12
<vagrantc>
yeah, i was CCed on part of it, at least
15:13
but our code actually supports the switch for /srv/tftp, it's just the documentation, no?
15:15
<alkisg>
Hmm I'm not really sure
15:15* alkisg looks...
15:17
<vbundi>
in /opt/ltsp/i386/lib/udev/ltspfs_entry it refers to /usr/share/ltsp/ltsp_config , is this file supposed to be missing?
15:18GodFather has joined #ltsp
15:20
<alkisg>
Heh... "/tftpboot was the old location, while /var/lib/tftpboot is default in Fedora 9+." - I wonder when people will settle on a specific hierarchy :)
15:21
The code seems fine...
15:21
<vagrantc>
vbundi: /opt/ltsp/i386/usr/share/ltsp/ltsp_config should definitely be present.
15:22
<vbundi>
vagrantc: doh yeah I actually meant /etc/ltspfs/ltspfsd.conf
15:23
<vagrantc>
vbundi: that's optional.
15:23
vbundi: it's largely there to support the rare cases where ltspfs is installed without the rest of LTSP
15:23
<vbundi>
oh ok
15:24
<vagrantc>
the code should check if it's there or not... right?
15:24
<vbundi>
vagrantc: I'm curious about this ltspfs_entry file because when I renamed it so that it wouldn't run my boot time sped up by 3 minutes
15:25
<vagrantc>
vbundi: it could probably use some serious re-writing.
15:25
vbundi: but it's what is called by udev to support usb sticks, cdroms, floppy drives, etc.
15:25
<vbundi>
vagrantc: yea it checks to see if it's there
15:26
vagrantc: ohh so by disabling this, it doesn't allow me to have local devices like that?
15:26
<vagrantc>
i did some ugly code to support usb CDROMs in there that is probably slowing it down considerably.
15:26
vbundi: yes, it was there for reasons other than slowing down your boot times :P
15:26
<Gadi>
time out
15:26
vagrantc: it isnt ur cDROM code
15:26
<vagrantc>
if i wanted to have it run slow, i'd just use sleep 300 or some such
15:27
Gadi: well, i *know* my cdrom code is ugly and probably could use some optimizations...
15:27
<Gadi>
we need to have ltsp_config write a cache file just for udev purposes
15:27
<Briareos1>
puh ... enough for today ... have a nice night / day, thanks & see you soon
15:27
<vbundi>
vagrantc: I wasn't assuming that you were trying to steal 3 minutes out of my day ;)
15:28
see ya
15:28
<stgraber>
vagrantc: feature freeze is next thursday, so ideally, I'd have 5.2 out before that
15:28
<vagrantc>
Gadi: it actually needs to check if they're running, though ... i.e. you unplug a usb cdrom, then caching values becomes wrong.
15:28rhodan has quit IRC
15:28
<Gadi>
not caching for your cdrom
15:28
caching for ltsp_config
15:29
what vbundi means is that if we do NOT source ltsp_config from ltspfs_enty it speeds up his boot times
15:29
<vagrantc>
ah.
15:29
<Gadi>
this is because udev makes calls in parallel
15:29Briareos1 has quit IRC
15:29
<vagrantc>
probably fairly race prone
15:29
<Gadi>
and ltsp_config was not written to be called in paralle
15:30
<vagrantc>
oh fun
15:30
<Gadi>
yeah
15:30
well, there's a quick fix
15:30
which is to make any lts.conf param affecting ltspfs_entry only set on boot
15:31
<johnny>
or just look at the environment vars?
15:31
<Gadi>
by means of having ltsp_config write a separate cache file which is sourced by ltspfs_entry if present INSTEAD of sourcing ltsp_config
15:31
<johnny>
/me is confused
15:31
how about we don't?
15:31
isn't there a way to just look at the environment vars?
15:31
and not source files except at startup? thus allowing them to be changed during runtime?
15:31frederickjh has left #ltsp
15:31
<Gadi>
multiple udevd's will have multiple env's I believe
15:32
<johnny>
sure.. but ther'es still only one system environment..
15:32
<Gadi>
huh? each shell has its own env
15:32
<johnny>
it starts a shell instance?
15:32
perhaps we need to store them somewhere else?
15:32
shm?
15:32
/proc ?
15:32
/sys ?
15:33
<Gadi>
well, we should sort it out somehow
15:33
<johnny>
sorry.. just suprised that this problem and similiar problems keep occuring
15:33
obviously sourcing the file isn't the way to do it
15:34
<Gadi>
well, ogra is happy to accept patches
15:34
:)
15:39wwx_ has quit IRC
15:41
<vagrantc>
Gadi: but wasn't vbundi's problem just a sheer number of parallel ltspfs calls?
15:42
<Gadi>
vagrantc: no, the issue is that there are so many ltspfs calls because ltspfs calls ltsp_config and ltsp_config is not returning
15:42
because all of the ltsp_config's step all over each other
15:42
<vagrantc>
ah
15:42
<Gadi>
when they do all the figuring out default vars and such
15:43
as well as writing to the filesystem
15:43
we do not have any file locking in there
15:43
<vagrantc>
we could also make it block ... though that's risky.
15:43mgariepy has quit IRC
15:43
<vagrantc>
basically, a file lock check ...
15:43
<Gadi>
yeah
15:44
at the minimum we should file lock check
15:44
<vagrantc>
while -f /var/run/ltspfs/lock ; do sleep 1 ; done touch /var/run/ltspfs/lock
15:44
<Gadi>
and do a better job of storing "decisions" that shouldn't change from boot time
15:44
<vagrantc>
though i start to get uncomfortable about caching things that actually could change ...
15:44
<knipwim>
johnny: can i commit this one?
15:44
http://bazaar.launchpad.net/~wimmuskee/ltsp/ltsp-gentoo-dev/revision/1496
15:44
<Gadi>
right
15:44
I agree
15:45
<vagrantc>
i.e. if we want high-availability thin clients, the root server could even change
15:45
<Gadi>
we should try not to cache things that change
15:46
<johnny>
why not store the vars in shm then?
15:46dro has quit IRC
15:46
<vagrantc>
also, i put in chroot calls that would source ltsp_config from the initramfs...
15:46
<johnny>
i just feel you guys are going about this the wrong way ...
15:46
i'm not sure why
15:46
it just feels wrong..
15:47
<vagrantc>
johnny: i don't see why storing them in shm will solve anything
15:47
<johnny>
perhaps we just need some api for getting vars..
15:47
like a function call..
15:47
but.. without instantiating a shell..
15:48
there's some missing piece of abstraction..
15:48
<vagrantc>
perhaps... getltscfg
15:48
<johnny>
that gets the whole thing
15:48
<alkisg>
gconf :P :D
15:48
<johnny>
i mean one at a time..
15:48
lol.. not gconf..
15:48
<vagrantc>
really, ltsp_config is just a workaround for the limitations of getltscfg ...
15:48
<johnny>
perhaps we should fix getltscfg instead..
15:48
<alkisg>
Don't we also want to *store* information in ltsp_config?
15:48
<Gadi>
I think it would go a long way just to clean up the code with the knowledge that it could be called in parallel
15:48
<alkisg>
getltscfg only *reads* info, it can't write it...
15:49
<johnny>
knipwim, sure
15:49
<Gadi>
ie: ban the use of pgrep
15:49
<vagrantc>
alkisg: and we can't fix it?
15:49
Gadi: what's so wrong with pgrep?
15:49
<alkisg>
vagrantc: it should be another tool... I don't think getltscfg fits the bill here
15:49
<vbundi>
I see quite a bit of pgrep going on in my bootchart durring the nbd_monitor
15:49
<vagrantc>
alkisg: sure, possibly... but another similar one.
15:50
<Gadi>
vagrantc: I think when multiple procs are calling pgrep at the same time, ur in for a world of hurt
15:50
<vagrantc>
Gadi: got it.
15:50
<alkisg>
And i'm not sure why shell isn't good enough already...
15:50
<johnny>
there's got to be a way to background some of this stuff .. in some always running process
15:50
something that makes it easy to call..
15:50
doesn't start a whole new process just to tell us something
15:50
perhaps i'm dreaming..
15:51
but there must be some way..
15:51shogunx has quit IRC
15:51
<Gadi>
most of the procs that we check for with pgrep *we* start ourselves
15:51jammcq has quit IRC
15:51
<Gadi>
quite easy to have a process touch a file to say that it's alive
15:52
<alkisg>
E.g.: ltsp_config checks a cache file, if it exists it just execs it. If not, it creates a lock file, generates the cache file, unlocks the lock, and execs the cache file...
15:52
<Gadi>
and remove it when it goes away
15:52
<vbundi>
I noticed quite a bit of it, even on my 2 minute bootchart http://imagebin.org/84115
15:52
<johnny>
sure.. but ...
15:52
<alkisg>
And if e.g. the server needs to be changed, then one just deletes that cache file
15:52
<vbundi>
down near the bottom
15:52paolino has quit IRC
15:53
<johnny>
why can't we do stuff in /etc/profile ..
15:53
or whatever we can do so that udev already knows stuff each time..
15:53
i don't know.. just braining..
15:53
/me realizes he has to get to work soon before the snow gets really crazy
15:53
snopocalypse 2
15:53
we're breakin a season record here
15:54
<vbundi>
where at
15:54
<johnny>
62 inches is the season record
15:54
we're at 60
15:54
half almost 2 feet on the ground
15:54
somewhere between 12 and 20 inches incoming
15:54
vbundi, baltimore, md
15:54
usa
15:54
<vbundi>
wow, hope your city cleans it's streets better than mine
15:54
<johnny>
not really
15:55
baltimore doesn't have the kind of setup as a city like chicago
15:55
or cities in other northern states
15:55
we're not prepared for this
15:55
as this is way out of the ordinary
15:59shogunx has joined #ltsp
16:01jhutchins_lt has quit IRC
16:01Gadi has left #ltsp
16:02alkisg has quit IRC
16:03spectra has quit IRC
16:04wwx has joined #ltsp
16:09mikkel has quit IRC
16:17etyack has quit IRC
16:19jhutchins_lt has joined #ltsp
16:32johnny has left #ltsp
16:43grantk has left #ltsp
17:03lucascoala has joined #ltsp
17:05johnny has joined #ltsp
17:08pmatulis has joined #ltsp
17:15bobby_C has quit IRC
17:17highvoltage has quit IRC
17:31waldo323 has joined #ltsp
17:32vagrantc has quit IRC
17:35waldo323 has quit IRC
17:35waldo323 has joined #ltsp
17:56highvoltage has joined #ltsp
18:02waldo323 has quit IRC
18:02waldo323 has joined #ltsp
18:14damianos has joined #ltsp
18:21mgariepy has joined #ltsp
18:28mgariepy has quit IRC
18:53damianos has left #ltsp
19:07lns_ has quit IRC
19:09lucascoala has quit IRC
19:11litlebuda has quit IRC
19:14lucascoala has joined #ltsp
19:15pmatulis has quit IRC
19:18cliebow has joined #ltsp
19:25litlebuda has joined #ltsp
19:28litlebuda has quit IRC
19:36pmatulis has joined #ltsp
19:38waldo323 has quit IRC
19:42lucascoala has quit IRC
20:03
<sbalneav>
Evening all
20:05ogra has quit IRC
20:16cliebow has quit IRC
20:27cliebow has joined #ltsp
20:39
<cliebow>
v
20:49daya has joined #ltsp
20:53cliebow has quit IRC
21:05johnny has left #ltsp
21:12Sarten-X2 has joined #ltsp
21:12Sarten-X has quit IRC
21:15GodFather has quit IRC
21:39Egyptian[Home] has joined #ltsp
21:54johnny has joined #ltsp
21:58Faithful has quit IRC
22:09jammcq has joined #ltsp
22:09
<jammcq>
hellooooo
22:13Faithful has joined #ltsp
22:28GGD has quit IRC
22:34pmatulis has quit IRC
22:40Lns has quit IRC
22:47johnny has left #ltsp
22:48johnny has joined #ltsp
22:54Lns has joined #ltsp
23:03johnny has quit IRC
23:08try2free has joined #ltsp
23:33johnny has joined #ltsp