|00:48||vagrantc has left IRC (vagrantc!~vagrant@unaffiliated/vagrantc, Quit: leaving)|
|01:19||GodFather has left IRC (GodFatherfirstname.lastname@example.org, Ping timeout: 240 seconds)|
|01:25||BassetFever has left IRC (BassetFever!a2f5d1e5@gateway/web/freenode/ip.184.108.40.206, Quit: Page closed)|
|02:30||andygraybeal has left IRC (email@example.com, Ping timeout: 276 seconds)|
|02:44||andygraybeal has joined IRC (firstname.lastname@example.org)|
|03:53||kjackal has joined IRC (email@example.com)|
|04:31||kjackal has left IRC (firstname.lastname@example.org, Ping timeout: 276 seconds)|
|06:26||semo163 has joined IRC (semo163!5f43d329@gateway/web/freenode/ip.220.127.116.11)|
alkisg: 2 news about nbd disconnects of my ubuntu ltsp5 pnp server and thin/fat clients
first one is very good for ltsp
actually all my disconnects were not because of some misconfiguration of ltsp staff, so it works as suppose to
What was the cause?
bad news : the reason was in cisco switch
Cisco switches manage to cause tcp timeouts? how is that possible?
i changed it just for those clients-server for dlink (cheapest smart one) and clienst stay alive for weekends now
i don't know in details about cisco switch and what is exactly in it cause disconnects
Ah ok its hardware might be broken then, it's not necessarily by design...
Nothing important to the ltsp world :)
so thank you alkisg for your and colleges hard work to get ltsp going
going to find out wat is causing disconnects in cisco switch next week
It'll probably be easier to do it without ltsp; just an nbd-server / nbd-client connection, an ssh connection etc, between a client and a server
So that you can see the logs etc without the clients hanging
if ssh stays up while nbd dies, that will indeed mean some software issue
to find out i/ve made 2 experiments^ started 2 fat (but diskless) clients one with autologin second with just nbd connection (without loggining) but both disconnects the same time after about 1hour 50 vin
in cisco i mean
man tcp => tcp_keepalive_time (integer; default: 7200; since Linux 2.2)
The number of seconds a connection needs to be idle before TCP begins sending out keep-alive probes. Keep-alives are sent only when the SO_KEEPALIVE socket option is enabled. The default value is 7200 seconds (2 hours). An idle connection is terminated after approximately an additional 11 minutes (9 probes an interval of 75 sec‐ onds apart) when keep-alive is enabled.
|06:43||yanu has left IRC (email@example.com, Remote host closed the connection)|
So the disconnection happens in 2 hours, 11 minutes, if it's a keepalive issue
when client boots it says block nbd9: Receive control failed (result -32) Apr 9 11:38:26 gnutova kernel: [ 31.696825] block nbd9: queue cleared
If cisco has a setting to block that, I don't think it would be disabled by default, it would be something in your setup
Ignore the nbd9 message, it's normal
and i think this cause 2 hours later disconnects from cisco which thinks this is dead session
im not expert in this though
i've changed TCP_KEEPALIVE kernel params as one german user suggested (i cannot find it now)net.ipv4.tcp_keepalive_intvl = 60 net.ipv4.tcp_keepalive_probes = 20 net.ipv4.tcp_keepalive_time = 600
they got cisco switch problems as well
but after they changed tcp_keepalive params as above problem disappeared
It's probably an unrelated forum message, ignore it
It doesn't match your symptoms
some archive of 2008
switch cisco SG300-52
switch dlink DES-1210-52
Have you tried a factory reset on the cisco switch?
no difference in logs with 2 different switches
i mean syslog on clients server (ltsp ones)
cisco actually brand new with just minor security settings like dhcp snooping (trusted) ip source guard and arp inspections
all interfaces on cisco correspond to hosts (i mean if dhcp server on say int gi01 it trusted as ltsp dnsmasq)
OK, it's completely beyond ltsp, maybe you should ask in #cisco then
on dlink no config but default and it works fine (i actually planned to replase it with cisco :-( )
|07:04||ricotz has joined IRC (ricotz!~ricotz@ubuntu/member/ricotz)|
yes I know it looks like cisco special question... just for info for others
semo163: I guess there is a default rule in Cisco device sthat says "disconnect idle connexions after n seconds", I have had the same problem for my SSH sessions to a specific remote server. A Cisco device, too.
It'd be nice to understand how that happens, since the switch is not an endpoint to the connection so it can't just "drop it"
it can act as a transparent firewall however
|07:42||semo163 has left IRC (semo163!5f43d329@gateway/web/freenode/ip.18.104.22.168, Ping timeout: 250 seconds)|
|07:43||semo163 has joined IRC (semo163!5f43ebf9@gateway/web/freenode/ip.22.214.171.124)|
And it would allow all the keepalive messages, but then block them, and even block reconnection attempts?
But i cannot such settings on SG300-52
mmm no. You are right.
The explanation "it completely blocks keepalive messages" sounds more sane...
actually I use this switch for ssh for all day long and there are lots of idle sessions for >5 hours, when I get back it stay alive
ok, so these are not the same symptoms
ssh has more advanced mechanism for keeping alive a connection afaik
something like http://unix.stackexchange.com/questions/34004/how-does-tcp-keepalive-work-in-ssh
semo163, it's also very easy to use aoe if you want, which won't have the keepalive issues as it operates on the ethernet level
aoe instead of nbd
point please to how to configure it
is there some resource to read on how to do so
Hmm I don't have any tutorial link handy, google a bit for aoe ltsp... it only requires 3-4 commands
ok. Thank you.
by the way i've got iscsi traff goes through this switch with no problems (no disconnects or so)
That too might not rely only on tcp-keepalive, it might use other means itself, like pinging
You can use a simple `nc` if you want to test tcp keepalive without additional protocols
|10:07||kjackal has joined IRC (firstname.lastname@example.org)|
|11:57||kjackal has left IRC (email@example.com, Ping timeout: 276 seconds)|
|12:24||schlady has joined IRC (firstname.lastname@example.org)|
|12:30||schlady has left IRC (email@example.com, Remote host closed the connection)|
|12:31||schlady has joined IRC (firstname.lastname@example.org)|
|14:07||andygraybeal has left IRC (email@example.com, Ping timeout: 252 seconds)|
|14:08||labkomltsp^labko has joined IRC (labkomltsp^labko!73b2d416@gateway/web/freenode/ip.126.96.36.199)|
i have a problem on my epoptes, any body can help mi
|14:16||tharkun has left IRC (tharkun!~0@unaffiliated/tharkun, Ping timeout: 250 seconds)|
|14:17||tharkun has joined IRC (firstname.lastname@example.org)|
|14:20||andygraybeal has joined IRC (email@example.com)|
|14:25||GodFather has joined IRC (GodFatherfirstname.lastname@example.org)|
|14:35||labkomltsp^labko has left IRC (labkomltsp^labko!73b2d416@gateway/web/freenode/ip.188.8.131.52, Ping timeout: 250 seconds)|
|14:56||schlady has left IRC (email@example.com, Read error: Connection reset by peer)|
|16:51||andygraybeal has left IRC (firstname.lastname@example.org, Read error: Connection reset by peer)|
|17:35||vagrantc has joined IRC (vagrantc!~vagrant@unaffiliated/vagrantc)|
|18:07||semo163 has left IRC (semo163!5f43ebf9@gateway/web/freenode/ip.184.108.40.206, Ping timeout: 250 seconds)|
|18:10||gehidore is now known as man|
|18:10||man is now known as gehidore|
|19:18||andygraybeal has joined IRC (email@example.com)|
|20:28||ricotz has left IRC (ricotz!~ricotz@ubuntu/member/ricotz, Quit: Leaving)|
|21:17||GodFather has left IRC (GodFatherfirstname.lastname@example.org, Ping timeout: 240 seconds)|
vagrantc: I'm thinking of adding a cleanup.d/script that installs ltsp-client if it's not already installed as part of the ltsp-update-image -c process,
an example usage is to have a stretch VM without ltsp in vbox, and run ltsp-update-image -c /path/to/vdi and create a squashfs out of it
It's not too intrusive, is it?
of course it'll only be activated if it operates on a cow system...
|21:44||alkisg is now known as alkisg_away|
alkisg_away: so it would only install it in the resulting image? ... there's a certain elegance to that
|22:07||GodFather has joined IRC (GodFatheremail@example.com)|
|22:44||sutula has left IRC (firstname.lastname@example.org, Ping timeout: 252 seconds)|
|22:46||sutula has joined IRC (email@example.com)|