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


Channel log from 30 August 2010   (all times are UTC)

00:03bobby_C has quit IRC
01:16lasersoft has quit IRC
01:25rjune has quit IRC
01:29Mava has quit IRC
02:18johnny has left #ltsp
02:20dobber has joined #ltsp
02:27gnunux has joined #ltsp
02:28
<gnunux>
hi
02:42bobby_C has joined #ltsp
03:21bobby_C has quit IRC
03:27wwx has quit IRC
03:28Gn1K has joined #ltsp
03:29Gn1K has left #ltsp
03:32wwx has joined #ltsp
04:08komunista has joined #ltsp
04:32komunista has left #ltsp
05:12Vorona has quit IRC
05:14cascade has joined #ltsp
05:56cascade has quit IRC
06:02shogunx has quit IRC
06:30pmatulis has joined #ltsp
06:31MorningSon has joined #ltsp
07:02gentgeen__ has quit IRC
07:52vvinet has quit IRC
07:54denisesball has left #ltsp
08:01shogunx has joined #ltsp
08:09grantk has joined #ltsp
08:10alexqwesa_ has joined #ltsp
08:10alexqwesa has quit IRC
08:11vvinet has joined #ltsp
08:15korcan has joined #ltsp
08:31alexqwesa_ has quit IRC
08:31alexqwesa_ has joined #ltsp
08:36mistik1 has quit IRC
09:05rjune has joined #ltsp
09:12shawnp0wers has joined #ltsp
09:16alexqwesa_ has quit IRC
09:19alexqwesa_ has joined #ltsp
09:29alexqwesa_ has quit IRC
09:29alexqwesa_ has joined #ltsp
09:39alexqwesa_ has quit IRC
09:42alexqwesa_ has joined #ltsp
09:42thunsucker has joined #ltsp
09:43rjune has quit IRC
09:45nubae has joined #ltsp
09:45
<nubae>
hi there
09:47
<Appiah>
heya
09:47
<nubae>
getting some problems with latest lucid update from hardy... when trying to login from a client, the server hangs trying to login the client, then says no connection from server
09:47
any ideas?
09:47
<Appiah>
ltsp-update-image
09:47
ltsp-update-sshkeys
09:47
<nubae>
Thought it would be sshkeys update, but did that to no avail
09:47
<Appiah>
=/
09:48
<nubae>
ok, so u thought same thing
09:48
<Appiah>
check /var/log/messages
09:48
<nubae>
I'll try update sshkeys and image again, should it be image first then keys, or vice versa?
09:49
<Appiah>
good question ...
09:51
<nubae>
in the logs I've got a failed kernel module loading module-x11-xsmp
09:52
not sure its relevant though
09:55
ok, in thin client command I get failed to unknown userid
09:55
which is real odd
10:02korcan has quit IRC
10:02
<nubae>
this must be sshkeys related but done several updates now sskeys first then image and vice versa
10:03
same login problem every time
10:03korcan has joined #ltsp
10:20gnunux has quit IRC
10:23Mava has joined #ltsp
10:25Kyle__ has joined #ltsp
10:25
<Kyle__>
Hello.
10:26
I figured out what was causing my login problem: my ssh keys. Now I have another problem. I can update my ssh keys, then update my image, and I still have the problem. If I ssh to the server as root from a workstation, then (of course) it works...until it's rebooted again.
10:29staffencasa has joined #ltsp
10:30
<Kyle__>
Anyone?
10:31nubae has quit IRC
10:50redbaritone has joined #ltsp
10:51dobber has quit IRC
10:54
<redbaritone>
I've spent all summer getting LTSP set up to serve a computer lab. Now, the school's new computer techs have (correctly) reorganized the network behind a firewall, and with one subnet/DHCP server. Would it be difficult to just remove DHCP, NAT redirection, etc. from the LTS server and point the new DHCP server to it? Or are there dependencies that I'm not aware of?
11:00Gadi has joined #ltsp
11:03
<thunsucker>
redbaritone: it should not be that difficult
11:04
<redbaritone>
thunsucker: Didn't think so. Thanks.
11:04
<thunsucker>
redbaritone: use the next-server feature in your dhcpd.conf and adjust the new dhcp server to hand out the pxe boot files and etc.
11:05vagrantc has joined #ltsp
11:17
<redbaritone>
thunsucker: So, let me get this straight. I don't completely turn off DHCP on my LTS server? I just turn off it's capability to assign IP addresses and point it to the new DHCP server with the next-server command?
11:18
<thunsucker>
redbaritone: you can turn it off or do next-server or both
11:23jhutchins_lt has joined #ltsp
11:47alkisg has joined #ltsp
11:48alkisg has joined #ltsp
12:05alkisg has quit IRC
12:09alkisg has joined #ltsp
12:13
<alkisg>
redbaritone: so you had the classic 2-nic setup as shown in the 3rd picture there: https://help.ubuntu.com/community/UbuntuLTSP/LTSPWiring and your techs thought it would be better to have a 1-nic setup as shown on the 2nd picture?
12:19mistik1 has joined #ltsp
12:22thunsucker_ has joined #ltsp
12:26thunsucker has quit IRC
12:30thunsucker_ has quit IRC
12:30thunsucker_ has joined #ltsp
12:31mistik1 has quit IRC
12:33
<Kyle__>
Ack :(
12:34
My clients don't apear to have the correct ssh keys, even though I just updated them then updated my image!
12:35alkisg has quit IRC
12:39mistik1 has joined #ltsp
12:42
<redbaritone>
alkisg: Yes, I have (had) the classic 2-NIC setup. But I'm the one assuming the one-NIC setup might be best, since I will no longer be controlling the network. Can the 2nd NIC still be useful?
12:45hawkI has joined #ltsp
12:47Techtronic has joined #ltsp
12:49Techtronic has left #ltsp
13:09
<grantk>
Hello, Anyone know where /usr/share/ltsp-cluster-control/Terminal/util/Node.php gets it's ltsp load balancing info from? I am getting an error in my logs that is thrown by that file and my clients do not get an application_server ip.
13:13
I can not tell if this is a configuration error on my master or the application server.
13:14
<Gadi>
grantk: can you wget -O- -q http://localhost:8008 on the lbserver?
13:14
<grantk>
Yes, it returns the ip of the application server
13:14
<Gadi>
(I mean get an IP without error from that)
13:15
sounds like app server is fine
13:16
<grantk>
Gadi: ok, I know last week I found that the /opt/ltsp/i386/etc/ltsp
13:16
/getltscfg-cluster.conf was not created automatically so I made it myself
13:16
http://ltsp.pastebin.com/UyvPcgGg
13:17
There is acopy of it, ohh, Minus the path
13:19
<Gadi>
looks fine to me
13:19
assuming that is the IP of the control center
13:19
<grantk>
yep, thats what I thought from documents I found
13:20
<Gadi>
what is the error?
13:20thunsucker_ has quit IRC
13:21
<grantk>
ltsp-directory[5871]: ,ip=10.0.2.91,mac=00:50:8B:34:66:1D,display=0.0,code=2, - cannot_set_status: This message is not set in text-en.phpThis is not a valid IP address : appServIp=0.0.0.0
13:22
ohh, and the information screen that pops up on the client before they get to the ldm shows the appserver ip as the master/lbserver ip
13:23
<Gadi>
and if you run getltscfg-cluster -a from a shell on the client?
13:24
<grantk>
CLUSTER_CONFIGURED="False"
13:24
export CLUSTER_CONFIGURED
13:31hawkI has quit IRC
13:39
<grantk>
does getltscfg-cluster.conf have to have any special permissions?
13:39
I have it at 644
13:59
<Gadi>
grantk: sorry had a ph call
14:00
<grantk>
np, I had to run off for a bit too!
14:01jhutchins_lt has quit IRC
14:02
<Gadi>
grantk: did you add the node in the control center?
14:02
(ie through the web interface)
14:02
<grantk>
no
14:03
I was thinking the /etc/ltsp/lbsconfig.xml was handling that
14:04
<Gadi>
grantk: try creating a group in the interface and adding the client to that group
14:04
I remember when I first set things up having to have stgraber step me through some of the weird things in the interface
14:05
it was one of those "click this don't ask" type things
14:05
:)
14:05
<grantk>
Gadi: is create group listed, for nodes I have a couple of find then a node properties section.
14:05* vagrantc waves to Gadi
14:05
<vagrantc>
Gadi: finally got that arm thin client booting
14:06
Gadi: apparently, we were using the wrong serial port speed
14:06
<grantk>
nvm, got a create now.
14:10hawkI has joined #ltsp
14:16
<grantk>
Gadi: Thanks, got me one step closer. I had to set up a node using the Control Center than the appserver ip was assigned no problem. It does not actually let a user log in for some reason but I am closer.
14:18
looks like ltsp-cluster-accuntmanager kills users, I noticed that from ssh but was not sure if it would be different using ldm
14:29
<Gadi>
vagrantc: cool
14:30
<grantk>
found problem with kill users - pam config problem....https://answers.launchpad.net/ltsp-cluster/+question/101648 fixed it right up.
14:30
<Gadi>
grantk: good - yeah, accountmanager is supposed to kill users off for various reasons - like if it thinks they are logged in elsewhere and such
14:30
I, too ha ve found it to be aggressive
14:30
:P
14:31
been meaning to talk to stgraber about that but haven't seen him in forever
14:31
<grantk>
ohh, I like that idea but hope those changes don't make that too troublesome.
15:08nutron has joined #ltsp
15:09alkisg has joined #ltsp
15:11pmatulis has quit IRC
15:12
<alkisg>
redbaritone: "But I'm the one assuming the one-NIC setup might be best, since I will no longer be controlling the network." ==> the 2 nic setup is much easier and safer for thin clients, as they are isolated from the rest of the network. So you control your own dhcp, your ip ranges and everything on your thin client subnet. But of course a 1-nic setup is also not a problem, you just need to configure your other dhcp server to point to the ltsp server.
15:15
<redbaritone>
alkisg: That's the problem. I don't want them to be isolated from the rest of the network anymore. They need to be able to run Windows, with client-server programs that access other Windows services on the network. They aren't going to be edubuntu clients ALL of the time.
15:16
<alkisg>
redbaritone: yup, same here, I went for the 1 nic setup too. I even preferred https://help.ubuntu.com/community/UbuntuLTSP/ProxyDHCP so that I didn't have to modify my other dhcp server (==router)
15:17
<redbaritone>
alkisg: Thanks for the link. I'll probably do that.
15:18agtilden has joined #ltsp
15:19
<redbaritone>
alkisg: The other advantage to doing this is that ANY computer in the school can boot Edubuntu. :-)
15:19
<alkisg>
redbaritone: that was possible before, too
15:19
You would just need to use the proxydhcp link above
15:20
(or modify your other dhcp server)
15:21
<redbaritone>
alkisg: Not the way I had it. I was given access to one 24-port switch. Now that the rest of the network has been consolidated, my LTS server is just one port in the storm (so to speak).
15:22
<alkisg>
redbaritone: if the clients can access ports 22, 69, 2000, and 9571 of your server, then they can boot from it. Of course if you had those blocked by a firewall, they couldn't.
15:24
<redbaritone>
alkisg: No, the switch was kept physically separate from the rest of the network. Our previous network guy was a dumbass.
15:24
<alkisg>
redbaritone: how did the server connect to the internet then?
15:25
Didn't the other nic have access to the rest of the network?
15:25
<redbaritone>
Separate external IP address -- the result of his previous dumbass decision.
15:26
<alkisg>
Ah :) Yeah then other clients on the network couldn't reach the ltsp server at all.
15:26
<redbaritone>
Correct!
15:27
Poor school administrators had no clue he was so stupid.
15:41mistik1 has quit IRC
15:48
<alkisg>
vagrantc, Gadi, stgraber, if I test nbd-server -c (copy on write) enough and it's working fine, would you object in ltsp-nbd being modified so that it doesn't use cow/bind-mounts at all? (to save RAM, prevent crashes when RAM is exhausted etc)
15:49
<vagrantc>
alkisg: i know the maintainer of nbd in debian recommends against it for speed reasons...
15:49
or maybe that is old news, though
15:49
<alkisg>
Hmmm I didn't see any problems with hdparm -tT...
15:50
<vagrantc>
maybe it's working well in recent versions...
15:50
<alkisg>
OK, I'll also do some benchmarks then. If all show to be fine, any other problems?
15:50
<vagrantc>
alkisg: there's also this: http://bugs.debian.org/470963
15:51
<alkisg>
Yeah we'd need a symlink in /tmp (or somewhere) to get around that
15:52
ln -s /opt/ltsp/images/i386.img /tmp/exported-image-xxxx
15:52
<Gadi>
does it create a cow image on the server disk or in server ram?
15:52
<vagrantc>
disk
15:52
<alkisg>
The disk, but I imagine that would be cached in RAM as appropriate
15:52
<Gadi>
and if the disk gets full?
15:53
<alkisg>
Hell?
15:53
:D
15:53
<Gadi>
sounds like it would also slow down IO
15:53
and increase network traffic
15:53
especially on your fatty fat clients
15:53
<alkisg>
It would be cached from the client side too
15:53
<vagrantc>
alkisg: i think it was one of those the more you write to it the slower it gets sort of things
15:53
<alkisg>
Ouch, that would be bad
15:54
I thought it uses sparse files, so I expected a constant speed there
15:54
<Gadi>
I don't see the benefit
15:54
<vagrantc>
it would be helpful with low-ram clients
15:54
<Gadi>
if you were after some sort of persistence, then yeah
15:55
"helpful"
15:55
<vagrantc>
that would be interesting too
15:55
<alkisg>
Gadi, right now when update-db or update-xapian-apt-index etc are ran on fat clients, they exhaust the client RAM. Having an e.g. 200 MB sparse file on the server sounds much better to me than wasting 200 MB of RAM on the client
15:55
<Gadi>
is that not what NBD_SWAP is for?
15:56
and if you have 100 users
15:56
thats.... 20GB of sparse files
15:56
<alkisg>
Why would nbd_swap be any better than nbd-server -c?
15:56
It would use the same disk on the server, same IO etc
15:56
<Gadi>
right
15:56
then why change?
15:57
<alkisg>
And it would avoid all the aufs/cow/tmpfs overhead
15:57
I believe it'll require less RAM on the client, and it will have better speed
15:57
<vagrantc>
(it would still have cow overhead)
15:57
<alkisg>
cow => on the server, not the client though
15:57* Gadi thinks the proof is in the pudding
15:57
<vagrantc>
nothing like testing
15:57
<Gadi>
mmm..... pudding
15:58
why cant we have bts somewhere with pudding?
15:58
<alkisg>
Sure, but before going to lengthy implementation + testing, I thought I'd ask first :)
15:58
<Gadi>
alkisg: I would follow you to the ends of the earth
15:58
<vagrantc>
alkisg: it wouldn't be too hard to implement ... not sure about testing
15:59
<alkisg>
Urm, I think Mark already went there, noone followed him... :-/
15:59* alkisg wants a ticket for a space bus :D
15:59
<vagrantc>
heck... it already supports nbd-proxy, direct nbd ... may as well add no aufs/tmpfs
15:59
<alkisg>
Heh
16:00
<vagrantc>
worst that happens is we can't read the code :)
16:00
and switch to live-initramfs ...
16:02
<alkisg>
main() { char *s="main() { char *s=%c%s%c; printf(s,34,s,34); }"; printf(s,34,s,34); }
16:04
I think a good test would be to boot 10 fat clients, run apt-get update && dist-upgrade on them with e.g. 200 MB worth of updates, and see how that affects everything
16:05mistik1 has joined #ltsp
16:06
<alkisg>
(and then try the same thing with nbd_swap)
16:09* alkisg fat client laptops are so fat that their lids can no longer be closed
16:14shawnp0wers has quit IRC
16:23
<grantk>
is it possible to have a lts.conf with ltsp-cluster?
16:24jhutchins_lt has joined #ltsp
16:29grantk has left #ltsp
16:36mistik1 has quit IRC
16:37mistik1 has joined #ltsp
16:46mistik1 has quit IRC
16:49agtilden has quit IRC
17:06agtilden has joined #ltsp
17:12shawnp0wers has joined #ltsp
17:19alkisg has quit IRC
17:29irule has joined #ltsp
17:31mistik1 has joined #ltsp
17:32redbaritone has quit IRC
17:35vagrantc has quit IRC
17:40Gadi has left #ltsp
18:02jhutchins_lt has quit IRC
18:07agtilden has quit IRC
18:12korcan has quit IRC
18:17pmatulis has joined #ltsp
18:26nutron has quit IRC
18:36staffencasa has quit IRC
18:42jhutchins_lt has joined #ltsp
19:02jhutchins_lt has quit IRC
19:04ogra has quit IRC
19:04ogra has joined #ltsp
19:19* Kyle__ sighs
19:19
<Kyle__>
Anyone awake?
19:20
I dunno if I've asked this before....but with fat clients, if you put startx instead of ldm in your lts.conf, will it work? What happens?
19:26shawnp0wers has quit IRC
19:37agtilden has joined #ltsp
19:46ogra_cmpc has quit IRC
19:58irule has quit IRC
20:00ogra_cmpc has joined #ltsp
20:01shogunx has quit IRC
20:06jconlon has joined #ltsp
20:11
<jconlon>
is using dnsmasq dificult with ltsp?
20:14
I have a small network, it was recommended that I try dnsmasq. I read on ubuntu's ltsp docs that dnsmasq could conflict with dhcp3 server, which I have been having problems with anyways.
20:15
dnsmasq was recommended to resolve a couple of internal hostnames btw
20:19
<Kyle__>
jconlon: dnsmasq? What are you intending to do with it?
20:20* Kyle__ finds it pretty easy to just configure a full blown DNS server.
20:20
<jconlon>
well, the ubuntu server is going to be the ltsp / dhcp server, but i have an application server that also provides an image to some of the clients
20:20
i am at wits end working with dhcpd.conf and some of my warehouse clients
20:21
but what i really wanted to do, was pass the information in the ltsp / dhcp servers' etc/host file to the dhcp clients
20:21
<Kyle__>
jconlon: My dhcpd.conf is perhaps overly simple...
20:21
<jconlon>
the image on the other server has to be able to resolve its servers hostname
20:22
<Kyle__>
Ahh. Humm.
20:22
<jconlon>
I have some very dificlut pxe clients
20:22
<Kyle__>
jconlon: Can the network in the warehouse talk to the name server under normal conditions?
20:22
<jconlon>
yes
20:22
<Kyle__>
Hum.
20:23
But when booted into ltsp they can't?
20:23* Kyle__ used to to have to use hosts files for warehouse machiens, due to their slow/crappy network...but tha'ts not going to work too well on boot...
20:23
<jconlon>
the ltsp server is just redirecting those clients to the application server to download a custom image...
20:23
<Kyle__>
Ah I see.
20:23
<jconlon>
that image refers to the host name of the server not its ip address
20:23
<Kyle__>
Can they boot OK? IE is the need for the name resolution booting or usage?
20:24* Kyle__ wonders if you can just toss a hosts file inside of that image...
20:24
<jconlon>
i wondered the same thing, my software provider is less than excited about my noob linux knowlege
20:24
they are very expensive
20:25
<Kyle__>
Ugh. I hate it when software providers dont' want to help...but then again I only found two or three ever that really did help :)
20:25
<jconlon>
lol!
20:25
<Kyle__>
jconlon: can you edit the image at all?
20:25
There's a reason I always suggest barracuda networks' products when a need comes up. They were really _really_ good to me.
20:26
<jconlon>
I've thought about trying, but honestly, I almost jumped off the roof, I can get a couple of hp thin clients to pick up the image
20:26
<Kyle__>
Symantic on the other hand might have well told me to go eff myself.
20:26
<jconlon>
ive editied my dhcpd.conf a hundred times maybe
20:26
lol
20:26
<Kyle__>
jconlon: Toss it up on pastebin. I've got a few minutes while this image is transfering....
20:27
<jconlon>
I wish i could, its at the office, I could really use the help
20:27
they obvioiusly need special parramaters of some sore
20:27
they are hp t5135, seems they use some older pxe bios
20:28
<Kyle__>
Ugh.
20:28
no proper remote access?
20:28
<jconlon>
there would be, but i'm replacing a dying windows server that used to do that
20:29
I'm actually not the it guy, i'm the owner, and biz sucks, so i gotta do this myself
20:29
<Kyle__>
Are you using a standard dhcpd?
20:29
Hehe. I know those days.
20:29
<jconlon>
i believe so, all i know of dhcp3 is what ive taught myself in the last 3 days...
20:30
<Kyle__>
Oh boy. You're doing pretty well, don't sweat it.
20:30
I'm tryign to remember if you can pass hosts info via dhcp.... I know you can do a _ton_ with it that most people don't...
20:31johnny has joined #ltsp
20:31
<jconlon>
I will research, someone just suggested dnsmasq as a simpler way to get this done, but I wasn't so sure if this was a good idea
20:32
<Kyle__>
jconlon: I'd just do a full blown dns server. My gut is telling me you're using ubuntu... am I right?
20:32
<jconlon>
but I did get a little excited I might could make those clients work with dnsmasq instead, but that doesn't really matter, I'm so mad, I'm going to replace the clients just out of principle
20:32
correct 10.04
20:33
documentation on bind just seemed so complicated to resolve 1 hostname, but it might be fun anyways
20:33
<Kyle__>
it's worth it.
20:34
<jconlon>
cool, thank you. Have you ever seen office space
20:34
<Kyle__>
Many many times
20:34
<jconlon>
the copier scene, that's what im gonna do to these thin clients
20:35
<Kyle__>
jconlon: first google for an ubuntu named tutorial, then paste it here, I'll take a peek and let you know if it's sensible/worth following.
20:35
Then, once you have it, I'll toss my config files up on pastebin so you can compare it to what you're seeing. Maybe both in conjunction will help?
20:36* Kyle__ has actually let users office-space equipment headded for the dump. It really makes them happy.
20:36
<jconlon>
lol
20:36
hold on
20:36
https://help.ubuntu.com/community/BIND9ServerHowto
20:37
btw what distro do you use?
20:38
<Kyle__>
For servers I generally use CentOS/Scientific linux (binary-compatible RHEL clones) for servers, ubuntu for clients, and since LTSP is most happy serving like to like, ubuntu-server for the LTSP.
20:39
Avoid that howto...it's well intentioned, but it will drive you nuts if this is your first time.
20:45
<jconlon>
good deal, it just looked too time consuming
20:49
<johnny>
dnsmas is better
20:49
don't use bind unless you need it
20:50
dnsmasq is so much so much so much better with ltsp :)
20:50
unless you're using bind with sql db
20:50
that's what i deployed
20:51
jconlon, so yeah.. use dnsmasq :)
20:51
Kyle__, don't suggest bind
20:51
it's badnews
20:53
unless you have thousands of clients
20:57agtilden has quit IRC
20:57pmatulis has quit IRC
20:58
<Kyle__>
johnny: Oooh, I love bind
20:58
johnny: It's nowhere near as complicated as people make it out to be.
20:59* Kyle__ never tried bind with sql...
21:00
<johnny>
i've setup bind myself
21:00
it's not that bad
21:00
but dnsmasq is superior
21:02irule has joined #ltsp
21:05* Kyle__ shrugs
21:05
<Kyle__>
I can't give any help with dnsmasq, as I've never set it up.
21:06
Call it an ulterior motive ;)
21:18
<jconlon>
thanks, I appreciate the help, but I must go now
21:18jconlon has quit IRC
21:18ThE-RiX has joined #ltsp
21:22RiXtEr has quit IRC
21:25RiXtEr has joined #ltsp
21:26mistik1 has quit IRC
21:29
<johnny>
Kyle__, dnsmasq is easy to setup, it's just adding entries to /etc/hosts
21:29
and turning on the dns server in the config
21:29ThE-RiX has quit IRC
21:30
<johnny>
of course you can point it to a different file, but by default it comes from there
21:30
Kyle__, i also use it to server tftp as well
21:33mistik1 has joined #ltsp
21:51irule has quit IRC
22:21korcan has joined #ltsp
22:26cyberorg has quit IRC
22:27cyberorg has joined #ltsp
23:50wwx has quit IRC