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


Channel log from 23 September 2019   (all times are UTC)

00:05
<vagrantc>
some...
00:05
oh well
04:27bluejaypop has joined IRC (bluejaypop!~BlueJay@201.175.202.197)
05:01lucas__ has left IRC (lucas__!~lucascast@177-185-139-206.isotelco.net.br, Remote host closed the connection)
05:01statler has joined IRC (statler!~Georg@p54897245.dip0.t-ipconnect.de)
06:04elenigioti has joined IRC (elenigioti!5b8ad6e4@static091138214228.access.hol.gr)
06:04
<elenigioti>
Καλημέρα σας. Άλκη έχεις χρόνο να λύσουμε το θέμα της εγκατάστασης στο laptop;
06:05
<alkisg>
Καλημέρα elenigioti, ναι αμέ, κάνε vnc:
06:05
!vnc-edide
06:05
<ltsp>
vnc-edide: To share your screen with me, open Epoptes → Help menu → Remote support → Host: srv1-dide.ioa.sch.gr, and click the Connect button
06:05
<alkisg>
!vnc-dide
06:05
<ltsp>
vnc-dide: To share your screen with me, run this: sudo apt-get --yes install x11vnc; x11vnc -connect srv1-dide.ioa.sch.gr - this is a reverse connection, it doesn't need port forwarding etc.
06:05
<alkisg>
Το δεύτερο
06:05
sudo apt install x11vnc; x11vnc -connect srv1-dide.ioa.sch.gr
06:05
Από το live cd
06:05
<elenigioti>
άλκη είμαι από windows. απ ότι θυμάμαι στο ubuntu δεν μου λειτουργεί το wifi
06:05
<alkisg>
Χρειάζεσαι καλώδιο για να δω
06:06
Δεν έχει LAN; Μόνο wifi;
06:06
<elenigioti>
χμ οκ δώσε μου 2 λπετά
06:06
<alkisg>
OK
06:10elenigioti has left IRC (elenigioti!5b8ad6e4@static091138214228.access.hol.gr, Ping timeout: 260 seconds)
06:37woernie has joined IRC (woernie!~werner@p50867A86.dip0.t-ipconnect.de)
06:44
<alkisg>
Latest news about the NFS speedup: https://github.com/ltsp/ltsp/issues/27#issuecomment-533976774
06:45
I expect that the first two lines in that table are the same for NBD too
06:45
I.e. for multiple clients, a similar workaround would greatly benefit NBD/SSHFS users
06:48
...meh, all those years we had double LAN traffic than what was necessary :/
06:51
<vagrantc>
sshfs too? wow.
06:51
alkisg: it'll make ltsp19 look that much better :)
06:51
<alkisg>
Yeah everything except nfs defaulted to 128 KB, which is too large readahead for networked file systems
06:52
NFS defaulted to 15 MB, which was why it was so laggy that I didn't use it in 12.04 and switched to sshfs
06:52
(pitty, had I known back then, we wouldn't even switch to sshfs :D)
06:53
*ever
06:56* alkisg wants to test if that will also speed up mksquashfs
06:57
<alkisg>
or some "nocache" variant...
06:57
Kernel caching needs 25% of a pentium 4 cpu; it's a whole lot just for book keeping
06:58
<vagrantc>
could build the squashfs with eatmydata
06:59
<alkisg>
Ah I'm talking about read caching, not write caching
06:59
That's the amazing part, that read caching can cause so much trouble
06:59
<fiesh>
alkisg: rw,vers=4.2,rsize=1048576,wsize=1048576,namlen=255,hard,proto=tcp,timeo=600,retrans=2,sec=sys,clientaddr=192.168.3.115,local_lock=none,addr=192.168.3.1
06:59
<alkisg>
fiesh: cat /sys/devices/virtual/bdi/*/read_ahead_kb
07:00
<vagrantc>
alkisg: curious
07:00
<alkisg>
If you see something other than 128 there, this device has a problem
07:00
Usually, 15360, the default NFS 15 MB readahead
07:00
<fiesh>
alkisg: firefox takes 9.5s to start over NFS here
07:00
<alkisg>
fiesh: and with the patches it might need 4s
07:00
<fiesh>
alkisg: 128 and 15360
07:01
<alkisg>
Right, try echo 128 > that file that has 15360
07:01
Or even better echo 4 > ...
07:01
<fiesh>
so 4 in all of them?
07:01
<alkisg>
No, just the nfs one, the one that had 15360
07:01
Well... if the other one is sshfs, you can try that one too
07:01
<fiesh>
oh ok all NFS ones, we have a bunch ;)
07:02
<alkisg>
The new nfs default will be 128, so just put everything to 128 and test,
07:03
while the *ltsp19* default for all involved file systems will be 4
07:03
Which then lowers the network traffic to less than half :O
07:04
<fiesh>
wow
07:04
less than 2s
07:04
fml
07:05
<alkisg>
:)
07:05
<fiesh>
with 4
07:05
I'll test 128
07:05
<alkisg>
For one client, it should be similar
07:05
But if you run multiple clients, it'll save time AFTER it saves network traffic
07:05
Which you can't easily reproduce with 1 client only
07:05
<fiesh>
yeah no difference noticable
07:05
with one client that is
07:06
the difference is crazy
07:06
<alkisg>
I really wonder how all those NFS maintainers etc didn't fix it for 10+ years :/
07:06
<fiesh>
yeah that's quite startling
07:07
when's the bug fix going to come officially?
07:07
cos setting up a workaround for this might be worthwhile...
07:07
<alkisg>
I guess in kernel 5.5 or something, while for ltsp19, in a day...
07:07
For ltsp5, it'll never make it
07:08
<fiesh>
wait so does setting the rsize and wsize have the same effect?
07:08
for then one could just change the fstab entries?
07:09
<alkisg>
Not exactly, it makes it a bit better, like 80% better
07:09
Because smaller packets are always used (bad) and the cache is still 15*rsize, which is still large
07:09
<fiesh>
hmm I guess I could also write a shell script that changes the read_ahead_kb values during start-up
07:10
<alkisg>
Right, also note that if you're using block devices (like ltsp uses loop devices for root), then you need it there too
07:10
Keep an eye open in the recent ltsp git commits, I'll commit something today or tomorrow
07:10
..that you could probably use
07:10
<fiesh>
I could just write a script that checks every read_ahead_kb value and sets those that are 15360 to 128?
07:11
<alkisg>
Yes, but that won't save the traffic of multiple clients
07:11
<fiesh>
hmm
07:11
<alkisg>
So it'll be "new nfs default good" but not "new ltsp19 default good"
07:11
<fiesh>
wait so there's more than the kernel patch?
07:11
<alkisg>
The kernel patch makes nfs the same as local file systems
07:11
It was waaaay worse
07:12
LTSP19 will try to make all network file systems use 4k instead of 128k
07:12
which should be a *lot* better for multiple clients
07:12
As it saves *more than half the network traffic*
07:12
<fiesh>
ah ok
07:17bluejaypop has left IRC (bluejaypop!~BlueJay@201.175.202.197, Ping timeout: 276 seconds)
07:22bluejaypop has joined IRC (bluejaypop!~BlueJay@201.175.202.197)
07:30ZAJDAN has joined IRC (ZAJDAN!~zdenek@77.48.149.75)
07:41
<alkisg>
FSTAB_NFS="server:/home/nfs /home nfs defauls,nolock 0 0"
07:41
Does anyone know if the "defaults" word has a meaning there, or it should be just "nolock"?
07:41ricotz has joined IRC (ricotz!~ricotz@ubuntu/member/ricotz)
07:46
<vagrantc>
defaults would be better than defauls
07:49
i have definitely seen behavior on some systems where leaving out defaults will change the behavior, but it's filesystem and/or device dependent
07:49
and it's been ages since i've experimented with that sort of thing...
07:49
i'd err on the side of leaving "defaults" in.
07:57
<alkisg>
OK, ty for the typo too, this was from the man page, people might have had a bad time if using copy/paste from there :)
07:58
<vagrantc>
heh :)
08:27vagrantc has left IRC (vagrantc!~vagrant@unaffiliated/vagrantc, Quit: leaving)
09:02bluejaypop has left IRC (bluejaypop!~BlueJay@201.175.202.197, Ping timeout: 265 seconds)
09:21gdi2k has joined IRC (gdi2k!~gdi2k@host86-185-211-13.range86-185.btcentralplus.com)
11:58Faith has joined IRC (Faith!~Paty_@unaffiliated/faith)
12:49lucascastro has joined IRC (lucascastro!~lucascast@177-185-139-206.isotelco.net.br)
12:55section1 has joined IRC (section1!~section1@178.33.109.106)
13:15woernie has left IRC (woernie!~werner@p50867A86.dip0.t-ipconnect.de, Remote host closed the connection)
14:00os_a has left IRC (os_a!~Thunderbi@195.112.116.22, Quit: os_a)
14:59mmarconm has joined IRC (mmarconm!~mmarconm@unaffiliated/mmarconm)
15:20fiesh has left IRC (fiesh!~fiesh@hq.wsoptics.de, Ping timeout: 268 seconds)
15:32fiesh has joined IRC (fiesh!~fiesh@hq.wsoptics.de)
15:44shored1 has left IRC (shored1!~shored@87-92-122-167.bb.dnainternet.fi, Read error: Connection reset by peer)
15:44shored has joined IRC (shored!~shored@87-92-122-167.bb.dnainternet.fi)
15:53bluejaypop has joined IRC (bluejaypop!~BlueJay@201.175.202.197)
16:09bluejaypop has left IRC (bluejaypop!~BlueJay@201.175.202.197, Quit: Leaving)
16:23vagrantc has joined IRC (vagrantc!~vagrant@unaffiliated/vagrantc)
17:16mmarconm has left IRC (mmarconm!~mmarconm@unaffiliated/mmarconm, Read error: Connection reset by peer)
18:37
<alkisg>
fiesh: https://github.com/ltsp/ltsp/blob/master/ltsp/common/ltsp/55-images.sh#L339
18:37
You call set_readahead either without parameters, or with a specific mount point, e.g. /home
18:37
Without any parameters it handles all NFS mounts
19:13meo has joined IRC (meo!~systemdju@unaffiliated/mikeseth)
19:13
<meo>
hello
19:43
<alkisg>
hello
19:46
<highvoltage>
howdy
19:47
<uumas>
hiya
19:53
<Hyperbyte>
hi
19:54
<vagrantc>
greetings
20:03
<fiesh>
alkisg: ah nice
20:03
alkisg: thanks
20:03shored has left IRC (shored!~shored@87-92-122-167.bb.dnainternet.fi, Read error: Connection reset by peer)
20:04shored has joined IRC (shored!~shored@87-92-122-167.bb.dnainternet.fi)
20:18section1 has left IRC (section1!~section1@178.33.109.106, Quit: Leaving)
20:34
<vagrantc>
alkisg: does ltsp19 have issues with "non-standard" shells?
20:44Faith has left IRC (Faith!~Paty_@unaffiliated/faith, Quit: Leaving)
21:00ricotz has left IRC (ricotz!~ricotz@ubuntu/member/ricotz, Quit: Leaving)