00:11 | ||cw has left IRC (||cw!~chrisw@97-87-137-194.dhcp.stls.mo.charter.com, Changing host) | |
00:11 | ||cw has joined IRC (||cw!~chrisw@unaffiliated/cw/x-1182934) | |
02:42 | vagrantc has joined IRC (vagrantc!~vagrant@unaffiliated/vagrantc) | |
04:33 | vagrantc has left IRC (vagrantc!~vagrant@unaffiliated/vagrantc, Quit: leaving) | |
06:58 | ricotz has joined IRC (ricotz!~ricotz@ubuntu/member/ricotz) | |
09:57 | markit has joined IRC (markit!~marco@88-149-234-116.v4.ngi.it) | |
10:05 | <alkisg> Now that SSD disks are becoming very cheap and very fast, it might be a good idea to store at least /home locally
| |
10:06 | Question: how important is the ability for the users to "work in any pc", and have their /home there? Would you miss it, even if local /home access was 10 times faster?
| |
10:06 | <markit> alkisg: the possibility to sit in any PC is very important to me
| |
10:07 | one of the main "selling point" in my 2 schools
| |
10:07 | with 20 pc and Gbit connection /home is already fast enough
| |
10:08 | <alkisg> markit: currently, it is; but have you tried windows 10 with hdd and with ssd?
| |
10:08 | With hdd, windows needs 5 minutes for the disk activity to settle down, to make the system more workable,
| |
10:08 | with ssd, that isn't noticable
| |
10:09 | <markit> I've ssd in the ltsp server
| |
10:09 | <alkisg> This is going to affect more apps, even linux ones, in the future, as all people will have ssd anddevelopers will rely on them
| |
10:09 | So, gigabit networking is 10 times slower than ssd
| |
10:09 | And if you multiply that with 10 clients, that's 100 times slower
| |
10:10 | <markit> oh, that's right, kde has some really I/O intensive boot garbage in it that developers just with local disk don't notice, but nfs makes them a nightmare
| |
10:10 | <alkisg> Is that speed loss acceptable, for the "work anywhere" benefit?
| |
10:10 | Currently it is NOT that bad; but it will become more and more noticable in the future
| |
10:10 | <markit> it is for me, since children can sit in any PC, and if a PC is broken can be easely replaced just plugging a "backup one" that is in every lab
| |
10:11 | <alkisg> Well if it was a windows pc, and it was broken, the student would go elsewhere, and just lose his documents
| |
10:11 | That'll be the same now
| |
10:11 | <markit> but /home is useful only for data (and they have nothing very big) and configuration (should be not that big either)
| |
10:12 | ltsp is better than windows because (also) of this
| |
10:12 | <alkisg> So you would accept 100 times slower than ssd, for this benefit, ok
| |
10:12 | <markit> if ltsp has local /home then the difference with Win10 is reduced
| |
10:12 | no one complained so far because of speed
| |
10:12 | <alkisg> Yes, so far, I'm talking about the future
| |
10:12 | <markit> but I've got a lot of praise because of affidability and flexibility
| |
10:13 | <alkisg> Another idea is that in the future, one will be able to buy a 64 GB usb stick with SSD-like speeds,
| |
10:13 | so that "local home" then will mean "move your usb elsewhere"
| |
10:13 | I.e. /home/username will be in the user's pocket, as a usb pendrive
| |
10:14 | This will combine both of the benefits, but we're not there yet
| |
10:14 | <markit> alkisg: future? 10 years from now? we have usb2 pc where we have "new ones"
| |
10:14 | <alkisg> usb2 can do 40 MB/sec, it's already faster than gigabit
| |
10:14 | <markit> btw, I want to try 18.04 installation this week ("pnp fashion), is the netplan stuff automatically solved by your scripts?
| |
10:14 | <alkisg> But not all usb2 drives can do 40 MB/sec
| |
10:15 | Ubuntu doesn't use netplan by default
| |
10:15 | <markit> really? and systemd DNS?
| |
10:15 | <alkisg> !install
| |
10:15 | <ltsp> install: http://wiki.ltsp.org/wiki/Installation/Ubuntu for Ubuntu, or http://wiki.ltsp.org/wiki/Installation for other distributions
| |
10:15 | <alkisg> I document the steps for systemd-resolvd there
| |
10:15 | <markit> maybe I tried something with 19.04 (not ltsp related) and cried loud...
| |
10:16 | <alkisg> It's basically ltsp-config dnsmasq, with or without --enable-dns
| |
10:16 | <markit> ok, so ltsp-config takes care of systemd dns, correct?
| |
10:16 | <alkisg> yes
| |
10:17 | By default, it does nothing
| |
10:17 | So systemd dns continues working
| |
10:17 | If you --enable-dns, then dnsmasq replaces systemd dns
| |
10:17 | <quinox> I like home being on the server, but I'm already tying people to specific machines because of VirtualBox images that we keep on harddrives
| |
10:19 | <alkisg> quinox: yup; also note that if you're already doing this, you can set an FSTAB lts.conf directive, which will mount the local drive as /home, so that even launching firefox etc will be faster
| |
10:19 | <markit> alkisg: and if I want to use /etc/network/interfaces to config wan and 2 BOND interfaces for clients, does it work "out of the box"?
| |
10:20 | <alkisg> markit: sure; but you can do this from network-manager too, so that teachers can see if the connection is alive or not
| |
10:20 | I think ubuntu mate still ships ifupdown, which is what handles /etc/network/interfaces
| |
10:20 | <markit> alkisg: my teachers can only mess up configs and connectivity, they are dumb in technology
| |
10:20 | <alkisg> I think some flavors don't ship it though
| |
10:21 | They can see if the cable is connected. They cannot edit configs if you don't allow them to.
| |
10:21 | So you're making things harder AND losing benefits
| |
10:21 | But sure, it'll work, go for it
| |
10:21 | <markit> even if they have sudo rights? I'm not very found on all the permission stuff :(
| |
10:21 | <quinox> that's an interesting idea... it would require making some changes to how we make backups, but having things be faster would be nice
| |
10:22 | I'll try it out on the next person that complains things aren't fast enough
| |
10:22 | <alkisg> quinox: more stable too; as ext4 is the default; while sshfs has issues with file locks, atomic renames etc, they sometimes cause issues
| |
10:22 | <quinox> yeah, for stability I've replaced SSHFS with NFS a long time ago
| |
10:23 | <alkisg> markit: yes, you can prevent sudoers from messing with network manager. You can't prevent them from running sudo rm -rf /etc/network/interfaces :P
| |
10:23 | <markit> quinox: do you use kde by chance?
| |
10:23 | <quinox> We do, yes
| |
10:23 | <alkisg> quinox: nfs3 or 4? Are you using authentication for mounts? How?
| |
10:23 | <quinox> no authentication :(
| |
10:24 | <alkisg> Heh, I'd go for that; but vagrantc would never accept it :P
| |
10:24 | <markit> quinox: I have a big I/O issue at startup, I had to set nfs /home *(rw,async,nohide,insecure,root_squash,no_subtree_check)
| |
10:24 | <quinox> I mount the complete /home on startup, and users are managed through LDAP so as long as you don't sudo you don't get access to other people's files
| |
10:24 | nearly all of my users are root on the LTSP server anyway, so it's okay
| |
10:25 | <alkisg> quinox: so for your use case, you wouldn't object if I transfered /etc/passwd and /etc/shadow over the network, right? :) This would allow you NOT to use ldap :D
| |
10:26 | <quinox> I wouldn't object, correct, but I'm happy with my LDAP thank you
| |
10:26 | <alkisg> Sure sure just survying for other users
| |
10:27 | <markit> quinox: kde at startup tries to find a lot of icons of various formats and puts in a cache file under /var/tmp/kdecache-USERNAME, did you solved it in some way to speed things up?
| |
10:27 | <quinox> I'm looking forward to the new LTSP, I'm gonna try to get it up and running with NFS4 and Kerberos
| |
10:27 | I believe we have some hack for that, yes, let me look it up...
| |
10:28 | <markit> alkisg: I've also the problem that with kubuntu 16.04 (and probably 18.04) epoptes video broadcast is broken... do you think you can give me some support to try to find why and, maybe, fix it?
| |
10:28 | <quinox> cat << EOF > /etc/X11/Xsession.d/10ltspMM_rm_phonondevicesrc
| |
10:28 | rm "$HOME/.kde/share/config/phonondevicesrc"
| |
10:28 | KDEVARTMP=$HOME/.cache
| |
10:29 | export KDEVARTMP
| |
10:29 | EOF
| |
10:29 | <markit> quinox: with NFS4 I had the problem that all home files were "nobody:nogroup"
| |
10:29 | quinox: lol that's MY hack!
| |
10:29 | <alkisg> markit: I probably fixed it already for epoptes 1.0, and you're using the old one, 0.5.10
| |
10:29 | <quinox> I also run this hack for KDE users: apt --yes remove print-manager
| |
10:29 | <markit> quinox: but is not for that issue
| |
10:30 | <quinox> KDE completely crashes, we tracked it to something it tries to do with a printer in our network
| |
10:30 | <markit> quinox: wich kubuntu? 18.04?
| |
10:31 | <quinox> yes, it started somewhere on 16.04, it's still happening with 18.04
| |
10:33 | <markit> mmm does not happen to me with 16.04, I'll have to check if print manager is installed
| |
10:33 | alkisg: what about pnp if I don't want a specific (or more) packages be in the client image? Like the "print-manager" above, or epoptes "server"?
| |
10:33 | <quinox> we found a bug report that has the same stacktrace, supposedly they fixed it upstream
| |
10:34 | <alkisg> Instead of removing it you could probably just use RM_SYSTEM_SERVICES=...
| |
10:34 | markit: the epoptes "server" auto-disables itself
| |
10:34 | for print manager, RM_SYSTEM_SERVICES
| |
10:34 | <quinox> idunno, I don't care for KDE, and my KDE users don't care for print-manager, so we'll just apply this fix and be done with it
| |
10:34 | <alkisg> For /usr/share/applications/*.desktop, a simple rm command in lts.conf
| |
10:35 | <quinox> I made some wrapper around ltsp-update-image to apply our custom hacks automatically, so it's effectively the same as RM_SYSTEM_SERVICES
| |
10:35 | <alkisg> You can also add snippets in /usr/share/ltsp/cleanup.d
| |
10:35 | <markit> alkisg: so for specific programs I need in lts.conf /usr/share/applications/MyProgram.desktop ? where and how put "rm" in lts.conf?
| |
10:36 | <alkisg> INIT_COMMAND_REMOVE_PLUMA_FROM_MENU="rm -f /usr/share/applications/pluma.desktop"
| |
10:36 | INIT_COMMAND_REMOVE_PLUMA="rm -f /usr/bin/pluma"
| |
10:36 | Things like that
| |
10:36 | I.e. just a few commands to be ran on ltsp client startups
| |
10:37 | Removing doesn't cost disk space, and it's simpler to maintain one image than many images
| |
10:37 | <markit> so the line MUST start with INIT_COMMAND or what?
| |
10:37 | <alkisg> !lts.conf
| |
10:37 | <ltsp> lts.conf: (#1) http://manpages.ubuntu.com/lts.conf, or (#2) lts.conf manpage is available in the ltsp-docs package
| |
10:38 | <alkisg> Yes, INIT_COMMAND_xxx means "run this command at startup"
| |
10:38 | xxx can be anything, it's for you to remember what it is
| |
10:39 | <markit> can't find INIT_COMMAND in http://manpages.ubuntu.com/lts.conf or am I blind?
| |
10:39 | <alkisg> Could be; we're not very good at documenting things :D
| |
10:39 | You could also use RCFILE_xxx
| |
10:40 | <markit> "use the source, Luke" fashion, I guess (when documentation is incomplete)
| |
10:42 | alkisg: thanks a lot as usual, I've to run now :) see you soon, I'll try screen broadcast again and let you know
| |
10:42 | markit has left IRC (markit!~marco@88-149-234-116.v4.ngi.it) | |
10:45 | <quinox> I agree with vagrant in that the default installation should be secure, but it has to be workable as well, and finding the sweet spot isn't always easy
| |
10:45 | <alkisg> I'm tempted to completely avoid ssh-based authentication, and sshfs
| |
10:46 | Those that use ldap/nfs4/kerberos etc will be secure,
| |
10:46 | and those that want plain nfs3 will be insecure :D
| |
10:46 | It's hard to simplify security, when distributions don't really endorse sshfs for /home/username and pamssh for authentication
| |
10:48 | And those that use local /home, will be as secure as local home is
| |
10:49 | <quinox> When someone can physically touch a PC he can do a lot of things already
| |
10:50 | on the other hand, users will get worms and cryptolockers and shit, so having passwords on things do still help
| |
10:50 | (and backups, obviously)
| |
10:50 | <alkisg> For the far future, I'd go with "come with your usb for /home/username, and have it encrypted"
| |
10:50 | and if they give the correct username/password that can decrypt it, they can login
| |
10:52 | <quinox> For customer projects when it comes to security I usually start by listing the thread model, what kind of attack(ers) do we want to deal with and which ones we don't (DDoS and NSA)
| |
10:52 | it helps to crystalize the discussion, instead of going back and forth about what could possibly go wrong
| |
10:53 | I like the sound of that
| |
10:55 | USB3 goes up to 5 Gbit/s? wow
| |
11:00 | <alkisg> And if you have 10 clients, that's 50 gbps :)
| |
11:00 | Hard to beat with a simple gigabit server...
| |
11:01 | usb 3.1 is 10 gbps, etc etc
| |
11:01 | <quinox> it sounds like a fantastic setup: Safe, easy to understand and build
| |
11:02 | <alkisg> And I think that's per controller, so one can use 2 usb sticks, one for caching root and one for /home
| |
11:02 | (one that belongs to the school and the other to the user)
| |
11:02 | That'll be like a dual-ssd setup, rather fast
| |
11:03 | Lunch time, later!
| |
11:03 | <quinox> bye
| |
11:07 | since you said it's for the far future, do you think it's a lot of work to build? Would be pretty sweet for LTSP6, during installation you ask the user if he wants LTSP to be (1) perfectly safe [bring your own USB drives], (2) perfectly safe [bring your own LDAP+Kerberos+NFS4], or (3) safe enough when you can sort of trust your users [LTSP will setup NFS3]
| |
11:11 | nickolai has joined IRC (nickolai!1f2b0d7f@gateway/web/freenode/ip.31.43.13.127) | |
11:12 | <nickolai> Hi everyone!
| |
11:12 | <quinox> hello
| |
11:13 | <nickolai> Alkisg is here&
| |
11:13 | ?
| |
11:14 | <quinox> he just left for lunch
| |
11:15 | 5 minutes ago
| |
11:15 | <nickolai> Can anyone help me with one question? Is it possible to make Skin of Windows on LTSP?
| |
11:16 | Im ready to pay for this
| |
11:17 | Like this one but over the LTSP https://github.com/B00merang-Project
| |
11:17 | <quinox> after logging in it's all standard linux
| |
11:18 | The login via LDM can be skinned, but I don't know how far you can go
| |
11:19 | funny repo :)
| |
11:19 | that you should - give it a shot!
| |
11:19 | (I don't know if you already have LTSP set up)
| |
11:19 | that you should -> that should work
| |
11:20 | <nickolai> The thing is to make user friendly interface (For users that used to usу Windows, bun only browser and Libero)
| |
11:20 | I му got LTSP set up
| |
11:20 | шму пще ше ыуееув гз
| |
11:21 | Sorry? misstype
| |
11:21 | <quinox> okay, so try it with 1 account to use fe the Win10 theme
| |
11:21 | <nickolai> Poor english(
| |
11:21 | <quinox> if that works there are ways of applying that change to all users
| |
11:22 | <nickolai> Can i ask someone more professional to do this and im ready to pay for this
| |
11:24 | There are a lot of things which i don't understand
| |
11:25 | <quinox> sure
| |
11:25 | i don't, but maybe alkisg
| |
11:25 | just keep hanging in the chat and he'll be back at some point :)
| |
11:26 | <nickolai> Oh thanks a lot)
| |
12:36 | Alkisg is here?
| |
13:09 | <alkisg> nickolai: back :)
| |
13:09 | nickolai: ok since it's about a job, let's talk in private messages
| |
16:25 | GodFather_ has joined IRC (GodFather_!~rcc@143.59.184.72) | |
16:26 | GodFather has joined IRC (GodFather!~rcc@143.59.184.72) | |
17:14 | <alkisg> !vq
| |
17:14 | <ltsp> Error: "vq" is not a valid command.
| |
17:25 | GodFather__ has joined IRC (GodFather__!~rcc@143.59.184.72) | |
17:27 | GodFather__ has left IRC (GodFather__!~rcc@143.59.184.72, Client Quit) | |
17:27 | GodFather_ has left IRC (GodFather_!~rcc@143.59.184.72, Quit: Ex-Chat) | |
17:28 | GodFather has left IRC (GodFather!~rcc@143.59.184.72, Remote host closed the connection) | |
17:28 | GodFather has joined IRC (GodFather!~rcc@143.59.184.72) | |
19:32 | woernie has joined IRC (woernie!~werner@p5B09D775.dip0.t-ipconnect.de) | |
20:19 | ricotz has left IRC (ricotz!~ricotz@ubuntu/member/ricotz, Remote host closed the connection) | |
20:29 | woernie has left IRC (woernie!~werner@p5B09D775.dip0.t-ipconnect.de, Remote host closed the connection) | |
21:03 | nehemiah has joined IRC (nehemiah!~nehemiah@156.19.21.242) | |
22:56 | adrianor1 is now known as adrianorg | |