Please note:
In working with alkisg in #ltsp, we've identified tcp keepalive values
in sysctl.conf that I had set:
net.ipv4.tcp_keepalive_time = 7200
net.ipv4.tcp_keepalive_intvl = 7200
net.ipv4.tcp_keepalive_probes = 6
Meaning my intervals are 14 hours, not 2. I'm resetting these to the
defaul
Here's a good synopsis of what I've seen. If I read previous posts
correctly, then these nbd processes should be killed after two hours.
Here are the results of my experiments:
r...@www:~# grep nbd /etc/hosts.allow
nbdrootd: ALL: keepalive
nbdswapd: ALL: keepalive
Client boots - here are serve
I was hoping to find out if this is something I can do or not, I of
course think I can do it, but do not realize exactly how. I have three
servers running ltsp, clients get assigned a server randomly. Sometimes
they crash(low memory thin clients) and when this happens they usually
have a firef
Michael,
I do have that same issue. I am going to give the exec a try that Alkis
was talking about. I appreciate the explanation by the way. Pretty
interesting to understand it a bit more.
I am not horribly worried about the swap files since they have been
moved to a swap server that has plen
The number of nbd server processes I see is greater than the number of
thin clients I have. I don't know that it's 2* greater though, as each
client gets one nbd-server for i386.img and one for the /tmp/tmp.??
swap file. I've set up some more testing for this morning to
illustrate what I'm saying.
> You should separate the machine setup from the user setup.
Yes, but LTSP involves both "machine setup" (think of scripts SCREEN_XX, or
xorg conf, that are configurable through ltsp.conf) and user setup (scripts
ldm/rc*).
Still, I have to admin that ltsp.conf is mainly about machine setup and l
Alkis Georgopoulos schrieb am 09. Aug 2010 um 10:05:48 CEST:
> Στις 09-08-2010, ημέρα Δευ, και ώρα 09:52 +0200, ο/η Helmut Lichtenberg
> έγραψε:
> > Hi Alkis,
> > thanks for your explanations. May I ask some more details?
>
> > Until now I imagined this connection somehow like opening a file over
Στις 09-08-2010, ημέρα Δευ, και ώρα 09:52 +0200, ο/η Helmut Lichtenberg
έγραψε:
> Hi Alkis,
> thanks for your explanations. May I ask some more details?
> Until now I imagined this connection somehow like opening a file over the
> network and getting something like a filehandle. I could not unders
Frédéric Grelot schrieb am 09. Aug 2010 um 09:30:36 CEST:
> > The number of nbd server processes is not related to the users logged
> > in, but to the machines connected.
>
> Well, maybe that could be an option of ltsp since ldm is quite light :
> active nbd/swap (launch client+swapon device) when
Στις 09-08-2010, ημέρα Δευ, και ώρα 09:30 +0200, ο/η Frédéric Grelot
έγραψε:
> Well, maybe that could be an option of ltsp since ldm is quite light :
> active nbd/swap (launch client+swapon device) when a user logs in, and
> disactive it (if possible) when no user is logged on (swapoff+close
> nbd
Hi Alkis,
thanks for your explanations. May I ask some more details?
Alkis Georgopoulos schrieb am 09. Aug 2010 um 09:31:16 CEST:
> In a 10.04 default installation:
> * the client runs nbd-proxy to connect to the server
> * and then nbd-client to connect to nbd-proxy locally
> * the server open
> The number of nbd server processes is not related to the users logged
> in, but
> to the machines connected.
Well, maybe that could be an option of ltsp since ldm is quite light : active
nbd/swap (launch client+swapon device) when a user logs in, and disactive it
(if possible) when no user is
Στις 09-08-2010, ημέρα Δευ, και ώρα 09:13 +0200, ο/η Helmut Lichtenberg
έγραψε:
> Alkis Georgopoulos schrieb am 06. Aug 2010 um 15:44:16 CEST:
> > You can put an extra `exec` at the last line of /usr/sbin/nbdrootd so
> > that no nbdrootd processes exist at all:
> > PATH=/usr/bin:/bin:$PATH exec nbd
grant schrieb am 06. Aug 2010 um 15:14:45 CEST:
> A client that runs displays around the office was also
> connected to that server at the time I killed the processes yesterday.
> It is still running fine, I did not search out it's process because I
> did not care if it needed to be restarted.
Michael Blinn schrieb am 06. Aug 2010 um 15:23:06 CEST:
> And even with the regression to the -27 kernel on my 8.04.4 servers, I
> have 25 nbd-server processes laying around with only 3 users logged in
> as of this morning.
The number of nbd server processes is not related to the users logged in,
15 matches
Mail list logo
