|00:51||gbaman has joined IRC (email@example.com)|
|01:41||gbaman has left IRC (firstname.lastname@example.org, )|
|01:50||lucascastro has joined IRC (email@example.com)|
|03:38||lucascastro has left IRC (firstname.lastname@example.org, Remote host closed the connection)|
|05:33||quinox has joined IRC (email@example.com)|
|05:35||quinox1 has left IRC (firstname.lastname@example.org, Ping timeout: 256 seconds)|
|05:46||jgee has left IRC (email@example.com, Ping timeout: 256 seconds)|
|05:58||Miguel905 has joined IRC (Miguel905firstname.lastname@example.org)|
THIS IS A FREENODE BREAKING NEWS ALERT!! Hitechcg AND opal ARE GOING AT IT RIGHT NOW WITH A LOT OF FIGHTING AND ARGUING WOW YOU DON'T WANT TO MISS THIS!! TYPE /JOIN ## TO SEE THE ACTION...AGAIN TYPE /JOIN ## TO SEE THE ACTION!!
quinox vsuojanen kjackal adrianorg ltsp ogra_ lee fiesh sbalneav TatankaT vlt ||cw highvoltage markus_e92 tek__ book`_ spectra pppingme PeperPots____ alexxtasi[m] alkisg Hyperbyte bcg zama bitchecker tarzeau dragon788 sutula Natureshadow coseus zamba
|05:58||Miguel905 has left IRC (Miguel905email@example.com, Client Quit)|
|06:36||Statler|Home has joined IRC (Statler|Home!~Georg@p5B30E62B.dip0.t-ipconnect.de)|
|06:40||ricotz has joined IRC (ricotz!~ricotz@ubuntu/member/ricotz)|
|07:04||pppingme has left IRC (pppingme!~pppingme@unaffiliated/pppingme, Ping timeout: 265 seconds)|
|07:14||pppingme has joined IRC (pppingme!~pppingme@unaffiliated/pppingme)|
alkisg: ok, so the clients (all of them) reliably lose network connectivity over night and don't even regain it
fiesh: did you try a VM client?
Or, a standalone client that is still up and you could check its logs?
|07:36||kjackal has left IRC (kjackal!~quassel@2a02:587:3116:5200:810e:5a68:91b1:33c6, Ping timeout: 246 seconds)|
I just rebooted it, hehe, the standalone client, but the problem is I cannot do anything any more. dmesg complains about the nfs server not responding any more (which is obvious given the network's down), and ping went to network unreachable
I have remote logging on with syslogd, but of course it doesn't work any more when the network is down
alkisg: but I can reproduce it over night :) what's the next step you suggest?
Yeah that's why the standalone client was the key
standalone and VM
And, thinking what can be the cause; e.g. a cleaning lady that pulls the switch plug to put the vacuumn cleaner :P
(I've actually had that issue, it took me a while to get the school to discover that :D)
fiesh: btw, do you also have old-style (NFS-booted) clients that get the same issue?
I.e. so that we make sure it's not related at all to our changes?
the thin clients do not have that issue
I haven't tried the NFS fat client
Aaah got one idea. What about dhclient, maybe it's asking for a new ip lease
Boot a client and check if dhclient is running
Normally ltsp clients don't renew their leases at all, but possibly the dracut initramfs has bad support for ip renewal
|08:01||kjackal has joined IRC (firstname.lastname@example.org)|
that's a great idea, will check in a bit
fiesh: if you get this again this night, I'll try to reproduce it here locally
so there is no dhclient running at all
on the fat client
however, also not on the thin client
so who's renewing the lease anyway?!
alkisg: is there a dhclient supposed to be running?
fiesh: if dhclient isn't running, I assume there's no lease renewal, so that's not the issue
but shouldn't the lease expire at some point?
Yes, and ltsp clients ignore that completely :D
at least the dhcp server might give away the ip address to some other device?
Usually dhcp servers ping before assigning a new address
So it's not an issue
What's your dhcp lease time?
If it's less than a day, do try increasing it to 2 days just for a test
hmm whatever the isc dhcpd default sto
but I guess it should be mainly up to the client
they request a lease time
https://www.isc.org/wp-content/uploads/2017/08/dhcp41conf.html => The default is 43200 seconds.
No, dhcp clients don't request a lease time
This is 12 hours, it matches...
yes, it does match
Try increasing default-lease-time 4 times
If it's that, we'll check how to properly solve it
ok, good idea!
Although netbooted clients had better get an infinite lease anyway :)
the question is if that makes any sense, I suppose the lease time exists to allow ip addresses to be renewable
I'd be fine with using link level ipv6 too, which wouldn't require dhcp
Well, if you have new clients all the time, then it does make sense to have them renewable, but if not, it's best to give ltsp clients an infinite lease
or actual ipv6, we have that too, but I found it surprisingly difficult to use it, because our router does the ipv6 advertisement, and then it has to cross the server, and go to the clients
There's no guarantee in the protocol that a client will keep its current ip after a lease renewal
So, the only official way to keep ltsp clients running, is an infinite lease
ipv6 support in ltsp isn't well tested
Check also dynamic-bootp-lease-cutoff date in the dhcpd link above
ah so it does assign infinite leases by default to bootp?!
so I suppose it's a bug that the fat clients' lease isn't infinite, whereas the thin clients' one is?
It's possible that dracut is doing a different dhcp request, but I don't think that's actually the case
I believe thin clients are just more tolerant than nbd
nfs and sshfs can survive even network cable unplugging, while nbd can't
but from the documentation of dynamic-bootp-lease-cutoff, the dhcp lease should already be infinite
but the network goes away and doesn't come back
I believe both thin and fat clients do NOT request bootp leases
oh I see
isn't that incorrect behavior?
bootp isn't the same as pxe
It's an older protocol with its own issues
I know, but the lease type might be correct, no?
It's possible that ltsp clients could fake being bootp clients, but I'm not sure if sysadmins would like that
And it would require support from dracut and initramfs-tools, outside the ltsp code base
One other test that we can do, is to netboot a fat client using nfs
I'll first go for the 48h lease
That nfs will still be booted using dracut, but since nfs is more tolerant than nbd, it may tell us if the problem is dracut or disconnections
well the ping that I ran here stayed dead
so the network didn't come back at all
You're pinging from the server to the client and the client doesn't reply? Yeah that suggests an "ifdown" event...
But it's not certain, it might be that the client has hanged due to the nbd file system being unavailable
I've been pinging the server from the client
and ping was still running, it just didn't get any responses and complained that the network was down
but I think you're most likely right that dracut is to blame for the change
I could even live with setting the default-lease-time of dhcpd to 2^31-1, 68 years would be fine with me
Let's try 48 hours first, it should be enough as a test
changed the setting and rebooted this client (and not the others, so we have a reference)
I think you'll need to restart dhcpd for the change to take effect, and then reboot the client
If you can also boot a client over dracut/nfs, that would give us valuable feedback too
|08:52||Statler_Office has joined IRC (Statler_Office!~Georg@gwrz3.lohn24.de)|
I'll do that too
the NFS is going to have a lease time of 48h too then though
or should I give it 12h?
I can restart dhcpd again and reset it
Both of those tests are useful (nfs on 12 or 48h), so whatever you like there
ok, I'll leave it at 48h then for now
to have fewer parameters
|10:31||Statler_Office has left IRC (Statler_Office!~Georg@gwrz3.lohn24.de, Ping timeout: 256 seconds)|
|11:16||Statler_Office has joined IRC (Statler_Office!~Georg@gwrz3.lohn24.de)|
|11:23||lee_ has joined IRC (email@example.com)|
|11:27||zamba has left IRC (firstname.lastname@example.org, *.net *.split)|
|11:28||zamba has joined IRC (email@example.com)|
|11:28||lee has left IRC (firstname.lastname@example.org, Ping timeout: 260 seconds)|
|11:38||kjackal has left IRC (email@example.com, Ping timeout: 246 seconds)|
|11:44||Faith has joined IRC (Faith!~Paty_@unaffiliated/faith)|
|11:47||Statler_Office has left IRC (Statler_Office!~Georg@gwrz3.lohn24.de, Ping timeout: 264 seconds)|
|12:00||Statler_Office has joined IRC (Statler_Office!~Georg@gwrz3.lohn24.de)|
|12:03||kjackal has joined IRC (kjackal!~quassel@2a02:587:3116:5200:810e:5a68:91b1:33c6)|
|12:23||lucascastro has joined IRC (firstname.lastname@example.org)|
|13:11||ricotz has left IRC (ricotz!~ricotz@ubuntu/member/ricotz, Quit: Leaving)|
|13:11||ricotz has joined IRC (ricotz!~ricotz@ubuntu/member/ricotz)|
|13:27||nehemiah has joined IRC (email@example.com)|
Is there any known problem with LTSP when the server is running nvidia proprietary drivers?
I get a bunch of 'GLX missing on display...' errors. The only thing I can see that's different from other setups is that the nvidia proprietary drivers are installed.
|13:38||ben_roose has joined IRC (firstname.lastname@example.org)|
you're using pnp right? do the clients have nv cards?
the server's graphics capabilities are irrelevant
I use ltsp-manager
The clients do not.
so where do you get the errors?
if the clients don't have glx, than that would be expected
The error occurred when starting evolution mail. Setting the server to use Nouveau driver actually solved the problem. No idea how that works. But I'm happy that it works now.
|14:41||nehemiah has left IRC (email@example.com)|
|14:41||nehemiah has joined IRC (firstname.lastname@example.org)|
|15:08||nehemiah has left IRC (email@example.com)|
|15:23||lucascastro has left IRC (firstname.lastname@example.org, Remote host closed the connection)|
|15:23||lucascastro has joined IRC (email@example.com)|
|15:38||sbalneav has left IRC (firstname.lastname@example.org, Ping timeout: 255 seconds)|
|15:39||nehemiah has joined IRC (email@example.com)|
|15:47||sbalneav has joined IRC (firstname.lastname@example.org)|
|16:03||nehemiah has left IRC (email@example.com)|
|16:18||kjackal has left IRC (kjackal!~quassel@2a02:587:3116:5200:810e:5a68:91b1:33c6, Remote host closed the connection)|
|16:21||kjackal has joined IRC (kjackal!~quassel@2a02:587:3116:5200:10e6:a89b:a4b0:9c3b)|
|16:22||Statler_Office has left IRC (Statler_Office!~Georg@gwrz3.lohn24.de, Remote host closed the connection)|
|16:22||Statler_Office has joined IRC (Statler_Office!~Georg@gwrz3.lohn24.de)|
|17:35||Statler_Office has left IRC (Statler_Office!~Georg@gwrz3.lohn24.de, Ping timeout: 240 seconds)|
|17:45||vagrantc has joined IRC (vagrantc!~vagrant@unaffiliated/vagrantc)|
|17:48||vagrantc has left IRC (vagrantc!~vagrant@unaffiliated/vagrantc, Client Quit)|
|17:48||Statler_Office has joined IRC (Statler_Office!~Georg@gwrz3.lohn24.de)|
|17:49||vagrantc has joined IRC (vagrantc!~vagrant@unaffiliated/vagrantc)|
|17:55||lucascastro has left IRC (firstname.lastname@example.org, Read error: Connection reset by peer)|
|17:55||lucascastro has joined IRC (email@example.com)|
he left already; the reason is that the nvidia binary driver replaced system libraries, and if you use that installation with other graphical drivers things break
the open source drivers are excellent for office work though, so I don't bother with the closed source anymore
|19:24||jgee has joined IRC (firstname.lastname@example.org)|
|19:29||Statler_Office has left IRC (Statler_Office!~Georg@gwrz3.lohn24.de, Remote host closed the connection)|
|20:08||Statler|Home has left IRC (Statler|Home!~Georg@p5B30E62B.dip0.t-ipconnect.de, Remote host closed the connection)|
|20:24||lucas_ has joined IRC (email@example.com)|
|20:24||lucascastro has left IRC (firstname.lastname@example.org, Read error: Connection reset by peer)|
|20:26||lucas_ has left IRC (email@example.com, Client Quit)|
|20:27||lucascastro has joined IRC (firstname.lastname@example.org)|
|20:39||Faith has left IRC (Faith!~Paty_@unaffiliated/faith, Quit: Leaving)|
|22:33||ricotz has left IRC (ricotz!~ricotz@ubuntu/member/ricotz, Quit: Leaving)|
|23:36||ben_roose has left IRC (email@example.com, Remote host closed the connection)|