|00:06||SmallR2002 has left IRC (SmallR2002email@example.com, Ping timeout: 245 seconds)|
|01:15||vagrantc has left IRC (vagrantc!~vagrant@freegeek/vagrantc, Quit: leaving)|
|01:38||PhoenixSTF has left IRC (PhoenixSTFfirstname.lastname@example.org, Remote host closed the connection)|
|02:35||hachque has left IRC (hachque!quassel@2600:3c01::f03c:91ff:fe96:5060, Remote host closed the connection)|
|02:44||andygraybeal has left IRC (email@example.com, Ping timeout: 244 seconds)|
|04:01||Phantomas has left IRC (Phantomas!~Phantomas@ubuntu/member/phantomas, Ping timeout: 255 seconds)|
|05:01||staffencasa has left IRC (firstname.lastname@example.org, Ping timeout: 260 seconds)|
|05:10||staffencasa has joined IRC (email@example.com)|
|05:33||vagrantc has joined IRC (firstname.lastname@example.org)|
|05:33||vagrantc has joined IRC (vagrantc!~vagrant@freegeek/vagrantc)|
|05:50||alkisg has joined IRC (alkisg!~alkisg@ubuntu/member/alkisg)|
|05:50||* alkisg wonders why we aren't using nbd-client to export localdevs to the server instead of ltspfs...|
This way we would care about the client supported file systems, we would be able to partition client usb sticks from the server, get the device free space info etc etc
|05:56||* alkisg wonders if CD burning would also work over NBD|
alkisg: autounmounting is a little trickier
Ah, you mean about the bug report?
With nbd-client, we'd only export whole devices, not partitions
The file system mounting part would be done on the server
Normally, with udisks etc
and if someone yanks the usb stick?
without it getting umounted?
Then nbd-client would timeout and cause an automatic unmount
With a proper implementation, it should be the same as if the usb stick was yanked on the server
alkisg: so, instead of the suid root stuff we do for ltspfs, we'd have suid stuff that mounts the NBD device and somehow marks it as for that user only?
Yes, chown user:user /dev/nbdu1234d1
in theory, ltspfs handles the autounmounting better than that, though
|06:04||* vagrantc also wonders how many /dev/nbdNN are typically enabled.|
udev would send a message about the block device being removed, so then all of ifs partitions etc would be unmounted
looks like 16 total...
We would created nodes with the suid wrapper script
it supports an arbitrary amount?
And then nbd-client -d /dev/device
but does the kernel interface support new ones?
I don't know about that...
$ modinfo nbd | grep max
parm: nbds_max:number of network block devices to initialize (default: 16) (int)
alkisg: so on moderately large servers to support localdevs, they'd need to manually configure that...
maybe there's a .d directory we could add
There is, /etc/modprobe.d
We could drop a file there
Remove the arbitrary 128 device limit for NBD. nbds_max can now be set to any number.
128 isn't too bad, although you might have users with 3 devices each
An internal disk, a internal cd rom, and a usb stick, yup
But on bigs setups that can be increased
Although then I'm guessing people would use load balancing + multiple servers
Btw, how well does ltspfs handle localapps writing to the same usb stick as apps on the serveR?
even two usb sticks...
alkisg: i think it's fine- it uses ltspfs to access the same files.
although you might have file locking issues, not sure.
one clobbering the other might be possible, if they were writing to the same file
With the nbd-client implementation, we could possibly use LOCALAPPS_EXTRA_MOUNTS=/media
That would include even devices mounted directly on the server, where the user has access...
alkisg: then you've got a round-trip for all the devices
|06:29||alkisg1 has joined IRC (alkisg1!~alkisg@ubuntu/member/alkisg)|
|06:29||alkisg has left IRC (alkisg!~alkisg@ubuntu/member/alkisg, Ping timeout: 252 seconds)|
|06:29||alkisg1 is now known as alkisg|
True, that would be a downside when localapps are used extensively
it sounds like an idea worth exploring, for sure
I think it would indeed offer benefits when localapps are not used... block devices vs a fuse-based file system...
But for localapps, fat clients etc, I think the best would be to let the system mount everything locally, and then export the already mounted file system as a directory to the server
(remoteapps could use that in the fat clients case)
vagrantc: would you object to a new lts.conf variable, ROOT_HASHED_PASSWD="value from mkpasswd -m sha-512" ?
For debugging reasons, of course...
So that people don't have to set a root password, rebuild the chroot, test, then unset the root password, rebuild the chroot
alkisg: so it would add a password hash to /etc/passwd ?
Yup, using usermod from init-ltsp.d
sounds reasonable to me.
of course, if lts.conf is working, you can just specify SCREEN_NN=shell
Yes but it's not exactly the same thing
E.g. you can't enter root credentials to policykit
And if you only want to troubleshoot one client, and don't want to bother with mac addresses, setting a hashed password globally sounds a bit more safe than enabling screen_02 globally
(that's minor though)
Hmm I did get a temporary "file not found" when trying to access a file in <ltsp-client>/media/user/MOUNT from a localapp though
vagrantc: Why is there an <ltsp-client>/media/root/MOUNT as well as an <ltsp-client>/media/user/MOUNT ?
alkisg: yeah, i see the logic for it.
alkisg: that's for rdesktop
Hmm that could be solved if we created a local "ltsp" user for kiosk, rdesktop etc, right?
if it's feasible to run them as such, sure.
dunno if rdesktop requires root ...
we should definitely put as little running as root as possible, if we can
I don't think it does, it runs fine from within sessions
Even DMs are not run as root nowadays
|07:02||* alkisg is still missing some basic info about ltspfs...|
What does /run/drives/MOUNT do?
Why a (empty!) vfat mount appeared there for just a few seconds?
How does ltspfsd access the device as a file system?
What does `ltspfs 127.0.0.1:/var/run/drives/VOLUME /tmp/.user-ltspfs/VOLUME` do? There's nothing in /tmp/.user* ...
|07:08||leio_ is now known as leio|
ls /tmp/.user-ltspfs/ is only transiently mounted on it's way to media
it gets mount --move'ed later
|07:10||* vagrantc had patches a long time ago to directly mount in /media|
vagrantc: but where is the actual vfat mount? I can't see it...
it dramatically reduced the amount of code run setuid, but it required that dirs get created in /media/ with write permission of the user.
Actually while trying a few minutes ago, even /dev/sdb (the usb stick) was lost! While the /media/user/VOLUME mount point was still working!
alkisg: i think that's in /var/run/drives/VOLUME
It's empty though
it unmounts after inactivity
and remounts if someone's asking for a mount
or asking for files in there.
having a shell parked in the /media/USER/FOO dir helps with that while debugging.
i think that's enough to keep it open
but i could be wrong
|07:13||* alkisg also put an ls; sleep 1 loop there|
Yup it was needed, just having a shell there was not enough
But why unmount the stick after a bit of inactivity? Wasn't a `sync` enough?
Thanks though I finally got some sense out of ltspfs :)
Having a shell parked in root@ltsp12:/run/drives/VOLUME is enough to stop the unmounting
alkisg: i don't understand about "Wasn't a `sync` enough?
vagrantc: unmounting the usb drive every 2 seconds and remounting it when more data is needed sounds a bit unnecessary to me
I'm guessing some bug related to unclean unmounts... otherwise why not unmount it only when requested?
many filesystems complain loudly if removed before unmounted
Even after "sync"? I didn't know that
I thought that `sync` meant "write the data, the journal and everything else related"
alkisg: also, things like usb sticks... there's no process for when the device is actually removed
i think sync also behaves a little different depending on your filesystem
So. If I'm not using LTSP. and I do: (1) insert stick, (2) write stuff, (3) sync, (4) unplug the stick without unmounting it
==> then my system or stick is not "clean"?
|07:22||cyberorg has joined IRC (cyberorg!~cyberorg@opensuse/member/Cyberorg)|
alkisg: i believe so, but am not positive
Gotcha, thanks again
If the unmount helper proves to work in most DEs, maybe it'll help there...
it won't likely have data corruption, so that might be ok
So that the user experience is the same both with and without LTSP
alkisg: yes, that's a good move.
if they happen to manually yank it, it'd be fine, but being able to unmount it woudl be nice.
alkisg: would someone be able to run the helper manually an unmount other people's stuff?
i guess no more than what they could already do
$ tail -n 1 /etc/mtab
ltspfs /media/user/SOTIRIA fuse.ltspfs rw,nosuid,nodev,user=user,uhelper=ltspfs 0 0
No, because of the "user=user" there
The helper will run as the user
Not as root, there's no need for suid there
|07:26||* vagrantc forgets which parts are setuid|
lbmount and fusermount are
So the user runs them, and while they do run as root, they both check the user name/id before allowing unmounts
OK so to sum up... things I was to see in ltspfs: (1) drop cdpinger, (2) support unmounts from nautilus, (3) run rdesktop, kiosk etc as a user so that the /media/root/VOLUME mount point is dropped
|07:44||komunista has joined IRC (email@example.com)|
sbalneav: could I also interest you in https://bugs.launchpad.net/ltsp/+bug/1094998 (i.e. drop cdpinger) whenever you have some time... :)
|08:35||alexqwesa__ has left IRC (firstname.lastname@example.org, Quit: Хана X'ам !!!)|
|09:00||alkisg1 has joined IRC (alkisg1!~alkisg@ubuntu/member/alkisg)|
|09:02||alkisg has left IRC (alkisg!~alkisg@ubuntu/member/alkisg, Ping timeout: 244 seconds)|
|09:09||vagrantc has left IRC (vagrantc!~vagrant@freegeek/vagrantc, Quit: leaving)|
|09:22||Phantomas has joined IRC (Phantomas!~Phantomas@ubuntu/member/phantomas)|
|09:23||alkisg1 is now known as alkisg|
|09:46||Phantomas has left IRC (Phantomas!~Phantomas@ubuntu/member/phantomas, Ping timeout: 245 seconds)|
|09:51||alkisg has left IRC (alkisg!~alkisg@ubuntu/member/alkisg, Quit: Leaving.)|
|10:00||Phantomas has joined IRC (Phantomas!~Phantomas@ubuntu/member/phantomas)|
|10:55||Phantomas has left IRC (Phantomas!~Phantomas@ubuntu/member/phantomas, Ping timeout: 265 seconds)|
|11:11||komunista has left IRC (email@example.com, Quit: Leaving.)|
|11:11||Phantomas has joined IRC (Phantomas!~Phantomas@ubuntu/member/phantomas)|
|11:18||komunista has joined IRC (firstname.lastname@example.org)|
|12:05||Phantomas has left IRC (Phantomas!~Phantomas@ubuntu/member/phantomas)|
|12:08||andygraybeal has joined IRC (email@example.com)|
|12:20||adrianorg_ has joined IRC (firstname.lastname@example.org)|
|13:20||adrianorg_ has left IRC (email@example.com, Ping timeout: 245 seconds)|
|13:33||Phantomas has joined IRC (Phantomas!~Phantomas@ubuntu/member/phantomas)|
|14:13||Phantomas has left IRC (Phantomas!~Phantomas@ubuntu/member/phantomas, Ping timeout: 245 seconds)|
|14:34||markit has joined IRC (firstname.lastname@example.org)|
|15:09||vmlintu has joined IRC (email@example.com)|
|15:10||komunista has left IRC (firstname.lastname@example.org, Quit: Leaving.)|
|16:01||vagrantc has joined IRC (email@example.com)|
|16:01||vagrantc has joined IRC (vagrantc!~vagrant@freegeek/vagrantc)|
|17:56||telex has left IRC (firstname.lastname@example.org, Remote host closed the connection)|
|17:58||telex has joined IRC (email@example.com)|
|19:03||adrianorg_ has joined IRC (firstname.lastname@example.org)|
|19:31||elias_a_ is now known as elias_a|
|20:47||markit has left IRC (email@example.com, )|
|20:56||Phantomas has joined IRC (Phantomas!~Phantomas@ubuntu/member/phantomas)|
|21:08||Parker955 has joined IRC (Parker955firstname.lastname@example.org)|
|21:22||vmlintu has left IRC (email@example.com, Read error: Operation timed out)|
|21:22||vmlintu has joined IRC (firstname.lastname@example.org)|
|21:23||Parker955 is now known as Parker955_Away|
|21:27||Parker955_Away is now known as Parker955|
|21:27||mighty-d has joined IRC (mighty-d!~instalado@unaffiliated/mighty-d)|
i have a huge problem
my clients are sending a nbd can not open nbd
and im not sure how to fix it
i moved nbd to inetd and i can see it is open on netstat
and can connect to it on por 2000
i checked the image path on the tcpd call and it is ok
can you please help me?
|21:29||vmlintu has left IRC (email@example.com, Ping timeout: 260 seconds)|
release: please mention the linux distro and release you're using :)
im using ubuntu 12.04
with a standard ltsp-server-standalone package
you should use nbd-server's configs
ok, so i should stop the service at inetd?
i changed it because i t wasnt working
let me do this change
gotta run, hopefull others can help
|21:32||vagrantc has left IRC (vagrantc!~vagrant@freegeek/vagrantc, Quit: leaving)|
|21:38||Parker955 is now known as Parker955_Away|
now im getting mounting /dev/nbd0 on /rofs failed
|22:04||Parker955_Away is now known as Parker955|
|22:29||Gadi has joined IRC (Gadifirstname.lastname@example.org)|
|22:36||Parker955 is now known as Parker955_Away|
please help me, im having a bad day with nbd...i dont know why i keep getting the can not open nbd: no such file or directory error
everything is default
|23:31||Phantomas has left IRC (Phantomas!~Phantomas@ubuntu/member/phantomas, Ping timeout: 255 seconds)|
|23:47||Phantomas has joined IRC (Phantomas!~Phantomas@ubuntu/member/phantomas)|
|23:59||alexqwesa__ has joined IRC (email@example.com)|