stgraber | LaserJock: hey, can you subscribe edubuntu-bugs to the ltspfs package ? We're already subscribed to ltsp and ldm but we're lacking ltspfs and so I missed a few bug reports :( | 01:12 |
---|---|---|
stgraber | LaserJock: btw, you and ollie are the only two admins, would it be possible to make me an admin as well ? | 01:12 |
LaserJock | stgraber: on it | 02:12 |
LaserJock | stgraber: I've added ltspfs but ogra will have to add you as an admin | 02:14 |
stgraber | LaserJock: thanks, will poke ogra later | 02:39 |
humbolt | Nubae: ?? | 12:20 |
humbolt | I am having trouble with your fatclient script. | 12:26 |
humbolt | ogra: Is Nubae real or fiction? | 12:38 |
humbolt | oh, I just fount his appearance in an IRC log. He is real. | 12:38 |
humbolt | How come I never catch him here. Even though he seems Austrian also. | 12:39 |
humbolt1 | for some reason LTSP does not seem to be able to connect to nbdrootd | 14:05 |
humbolt1 | All necessary config is there, all services running, hosts.allow entries there, ... | 14:06 |
humbolt1 | any ideas | 14:07 |
humbolt1 | or is nbd not fit for xinetd? | 14:07 |
humbolt1 | what could be wrong, if my thin client boots up, but login does not work? | 18:23 |
=== sancas_ is now known as sancas | ||
humbolt | anybody here for a change? | 20:04 |
humbolt | LDM_DIRECTX is off by default I suppose? But does this only concern LDM or the following X session also?\ | 20:06 |
humbolt | same question for NETWORK_COMPRESSION, what is the default and does this concern the whole X session or just LDM? | 20:07 |
humbolt | ogra: could you help me with this? | 20:07 |
stgraber | by default everything goes through ssh and nothing is compressed | 20:08 |
humbolt | stgraber: alright. if I have an atom client, ergo enough cpu, which settings would lead to the most responsive desktop? | 20:10 |
humbolt | and the best video playback performance? | 20:10 |
humbolt | stgraber: can it be that LDM_DIRECTX = false actually runs faster on my client? | 20:11 |
humbolt | it felt like it | 20:11 |
stgraber | LDM_DIRECTX=true | 20:11 |
stgraber | no, it can't be faster with = false | 20:11 |
humbolt | stgraber: ok | 20:12 |
stgraber | as X in SSH is a lot slower than directly sent from the server to the client | 20:12 |
humbolt | stgraber: when LDM_DIRECTX=true, do I have to enable XDMCP in gdm? | 20:12 |
stgraber | no | 20:12 |
humbolt | ?stgraber: so how is the connection made then, if not tunneled through ssh | 20:13 |
stgraber | LTSP doesn't even need GDM on the server (unless you wantt a local session too) | 20:13 |
humbolt | stgraber: really, interesting\ | 20:13 |
humbolt | this damn keyboard | 20:13 |
humbolt | stgraber: but that only works, if not SCREEN_07=startx | 20:14 |
humbolt | right? | 20:14 |
humbolt | then it is an xdmcp session | 20:14 |
stgraber | indeed, you'll need gdm if you want xdmcp | 20:15 |
stgraber | but usually you don't want xdmcp as I'm pretty sure some things will break | 20:15 |
humbolt | NETWORK_COMPRESSION does only concern the ssh tunnel, right. ergo nondirectx | 20:15 |
stgraber | indeed | 20:15 |
humbolt | stgraber: and is only recommended for slow connections? | 20:16 |
humbolt | on by default? | 20:16 |
humbolt | stgraber: up to now, I don't see no usb pendrives plugged into my client coming up on the desktop. any idea? | 20:20 |
stgraber | is your user in the fuse group ? | 20:21 |
humbolt | yes | 20:22 |
humbolt | ok, actually I see 2 ltspfs processes on the server | 20:23 |
humbolt | ok, they are there | 20:23 |
humbolt | just don't show up on the desktop | 20:24 |
humbolt | I probably have that deactivated | 20:24 |
humbolt | is network compression on or off by default | 20:24 |
stgraber | off | 20:25 |
humbolt | works pretty well these days | 20:25 |
humbolt | especially with an eee box as client | 20:25 |
humbolt | have you ever tried nubae's fatclient setup? | 20:25 |
stgraber | nope, he uses NFS and I can't use it on the kind of networks I have (several thousands thin clients) | 20:26 |
humbolt | really! | 20:26 |
stgraber | so I work on getting the localapps working perfectly | 20:26 |
humbolt | several 1000s | 20:27 |
stgraber | yes | 20:27 |
humbolt | company or educational network? | 20:27 |
stgraber | I have one school district with 3k thin clients, another with 1k and a few more with pre-projects or ongoing deployments | 20:27 |
humbolt | where are you located? | 20:27 |
stgraber | I work for a company that does LTSP deployment in school and entreprises in the canada and US | 20:28 |
stgraber | we're in Sherbrooke, Quebec, Canada | 20:28 |
humbolt | quebec, great! | 20:28 |
stgraber | yep funny place, moved there 6-7 months ago (coming from Switzerland) | 20:29 |
humbolt | na servas | 20:29 |
humbolt | austria | 20:29 |
stgraber | hehe | 20:29 |
humbolt | was there for 6 month | 20:29 |
humbolt | 2 years back | 20:29 |
humbolt | Loved the fact, that they actually have some culture there compared to the US | 20:30 |
humbolt | and why exactly is nfs so bad for this kind of setup? | 20:30 |
humbolt | are we talking for home dirs or root? | 20:31 |
stgraber | both, NFS just ten to destroy my networks :) | 20:32 |
stgraber | We're doing centralized deployment so no server in the schools but everything in a datacenter | 20:32 |
humbolt | what kind of connection from the datacenter to the schools? | 20:33 |
stgraber | NBD can be compressed and doesn't trigger a read action everytime you see a file | 20:33 |
stgraber | fiber, between 100MB and 10GB | 20:33 |
humbolt | I see | 20:33 |
stgraber | with NBD you get caching on the thin clients and that means a lot less load on the server | 20:33 |
humbolt | alright | 20:33 |
stgraber | with the NFS our NFS server used to have a load between 100 and 120 with 3k thin clients | 20:34 |
stgraber | now the NBD has a 3-4 load when they're booting in the morning | 20:34 |
humbolt | I see | 20:34 |
humbolt | what are you paying for the 100mb to 10gb fiber? | 20:35 |
stgraber | no idea :) my company does the thin client deployment we don't do the whole networking | 20:35 |
humbolt | I see | 20:35 |
stgraber | usually school districts have their own infrastructure so it just cost them a lot to install it the first time | 20:36 |
stgraber | then it's their own | 20:36 |
humbolt | hmm | 20:36 |
humbolt | what kind of clients are you using there? | 20:36 |
humbolt | old pcs or something cool? | 20:37 |
stgraber | wyse thin clients (VIA based), webdt (Geode based), a few old computers and we're now working on finding a good atom base thin client | 20:37 |
stgraber | we're testing the eee box and thin clients from disklessworkstation | 20:38 |
stgraber | also the FX160 from DELL (don't like it, only the CPU is intel, everything else is SIS) | 20:38 |
humbolt | I am just now testing the eee box. So far so good. But I actually was thinking about moving to fat clients based on this hardware. | 20:41 |
humbolt | Still there are many things I am still concerned about with that. | 20:41 |
humbolt | And now that you mentioned the nfs issue, there is one more | 20:42 |
humbolt | am I understanding that correctly, with nbd temp files and logs written to var are stored in ram and lost on reconnect. | 20:43 |
humbolt | stored in the servers or the clients ram? | 20:43 |
stgraber | yes, but that's the same with NFS | 20:44 |
stgraber | the difference is that NFS exports a filesystem and so the hdd cache is done by the server | 20:44 |
stgraber | nbd exports a block device, so the cache is done on the thin client | 20:44 |
stgraber | reducing the number of read access done by the thin client (binary called very often or permission checks) | 20:45 |
humbolt | the cache. but what about newly created files | 20:45 |
stgraber | newly created files are in both cases (nbd and nfs) stored in a ramfs | 20:45 |
humbolt | so for a nbd based /home dir we'd need a cluster fs, so several clients can access the same blockdevice | 20:46 |
stgraber | no, we'd just do the same as for localapps and mount /home using sshfs | 20:48 |
humbolt | why is that better than nfs ... because fuse provides the caching? | 20:49 |
stgraber | because that way anyone on the network can't access your data :) | 20:50 |
humbolt | ok | 20:50 |
stgraber | it's extremely easy to get access to one's data with NFS | 20:50 |
stgraber | for the thin client root the main issue with NFS was the speed and CPU load on the server, for that the main issue is security | 20:51 |
humbolt | I see | 20:51 |
stgraber | with NFS everyone can read everyone's data without much trouble (IP spoofing in the worst case but that's easy on a LAN) | 20:51 |
humbolt | How good are local apps working these days? | 20:51 |
stgraber | not bad, I've my first deployment with firefox as localapp at the moment, still a few glitches when playing video (mplayer-plugin's fault) and I'll need an easy way of opening documents from the localapp firefox in a remote application on the server | 20:53 |
humbolt | wouldn't a fatclient be much easier to setup and handle in the long run? | 20:54 |
humbolt | and what about authentication in your domain? | 20:54 |
humbolt | you do ldap or aaa? | 20:55 |
humbolt | as you probably have a load balanced server env | 20:55 |
stgraber | ldap | 20:56 |
stgraber | fat clients would be a pain to maintain as you can't easily switch (as in on the fly) from a thin to fat client profile, updates would be a lot harder to push (basically reboot the whole network) and you'd need to let everyone on the network access the file server, ldap server, ... | 20:57 |
stgraber | at the moment the LAN only has access to the application servers and the application servers have access to the rest of the network | 20:57 |
humbolt | I see. So what are you using as file servers? | 20:59 |
humbolt | I was going to say, what if the root fs was not an image but a real block device, but I understand the benefit of the image lies in the squashfs | 21:00 |
stgraber | file servers are only used for /home on the application server, it depends on the customer but we do ncpfs (novell), cifs (novell/windows/linux) or nfs (but it's then limited to the application server and not available on the whole network) | 21:01 |
humbolt | I see | 21:02 |
humbolt | cifs, have not tried to mount one of these for a long time | 21:02 |
humbolt | smbmount used to be buggy with codepage handling. what do you use to mount cifs nowadays, and how are foreign codepages handled. does that work finally | 21:03 |
humbolt | I always had my mp3 filenames screwed up back in the days in university | 21:03 |
stgraber | seems to work quite well with recent cifs (cifs module in the kernel and mount.cifs or something similar) | 21:04 |
humbolt | you mount the individual home dir on login via pam or the whole thing from the fileserver to the ldm server? | 21:05 |
stgraber | individually | 21:06 |
stgraber | as they're often on different servers | 21:06 |
humbolt | Is your company hiring? | 21:07 |
stgraber | not sure, it's possible we still do (looking at the number of upcoming projects) | 21:08 |
humbolt | I'll paste you my email, if you ever need more hands and brains. I am just maintaining a smaller LTSP environment at the moment, but I am doing lots of other stuff with linux professionally. Used to work for a provider. | 21:09 |
humbolt | The one thing, I am having the most trouble with in my setup, is people just turning off the machine leaving running processes behind. | 21:11 |
humbolt | GNOME does not really take care of that. | 21:11 |
humbolt | What could I do to get this sorted? | 21:12 |
stgraber | I've written a daemon for that for ltsp-cluster | 21:12 |
stgraber | we can know where someone is connected and do the cleanup behind them | 21:12 |
stgraber | I believe there's also a simpler tool available done by another LTSP guy | 21:12 |
humbolt | I once installed an app for that | 21:13 |
humbolt | Did help a bit, but did not work all time. | 21:13 |
tom1 | greetings.. new user here. I installed ubuntu just fine but my hard drive confg didn't come out how I wanted. Basically I have a 25gb primary partition(which is want I wanted) but the rest of my drive is not being used and I can't figure out how to access it and configure as an extended drive(The free space is missing or unmounted). Are there any disk management tools I can use to config this? | 21:14 |
humbolt | Or are you using KDE? | 21:14 |
humbolt | tom1: gparted | 21:15 |
tom1 | i need to add the app gparted? | 21:15 |
humbolt | tom1: you must know yourself, I have no idea if it is installed on your system. | 21:16 |
tom1 | thanks for the help | 21:16 |
humbolt | tom1: sudo apt-get install gparted | 21:18 |
humbolt | stgraber: the gnome-watchdog is checking if gnome-panel is still running to determine if a user is not logged in anymore. Is your daemon doing the same thing? | 21:30 |
humbolt | back in a sec | 21:34 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!