Thanks for the very fast answer Roland.

I did learn tmux on my linux workstation, beside the fact that its really cool and plan to use it the future, I am still struggling to have the demo and fe working in tmux

When starting the demo-system (demo-system -a start) -> console-demo-vms, I know as a fact now that I am in tmux but I am still on the c1-head. any Ctrlr-t n (or p) said "no next windows"  , not only I can't switch to demo but also fe which  if I understand correctly must have started automatically.

Once I exit from tmux with Ctrl-t d,

tmux ls gives:
tmux ls
error connecting to /tmp/tmux-0/default (No such file or directory)

I don't see demo or fe.

When trying to start console-fe-vm
console-fe-vm
No tmux session 'login-vm' with server socket 'ql-vms'

What's happening here ?
note that during the install I rename the host c1-head jcluster and the fe jclustert   (jcluster and jclustert are found in the DNS).
Would it change anything ?

I I do
0  root@jcluster  ~  #
dsh -a uptime
beo-201: ssh: connect to host beo-201 port 22: No route to host
beo-202: ssh: connect to host beo-202 port 22: No route to host
beo-203: ssh: connect to host beo-203 port 22: No route to host
beo-204: ssh: connect to host beo-204 port 22: No route to host
login-c: ssh: connect to host login-c port 22: No route to host

So yes , no demo or fe connection.
Why do you thing is happening.


Thanks for your help




On 11/11/23 03:59, Roland Fehrenbacher wrote:
Hi Christophe,

On 11/11/23 07:28, Christophe Guilbert wrote:
Hi ,
I am completely new with qlustar. I would need some help to start having the cluster running as I don't really understand the doc.
I did install qlustar (4 times) with no problem , Since I did not find out how to run  the virtual FE node (and the demo mode VM by the way ). The last install was without create a virtual  FE node.
Following the section "1.5.5. Spack Package Manager" , it assume that I installed the FE node which I did not . There' s no alternative.
I tried in vain to  switch from root to user softadm in the main terminal , the user softadm is not even in the /etc/passwd.
Under root I try my luck with the spack command leading to :
spack compiler find
-bash: spack: command not found

How to do the spack section without creating the virtual FE node ?

you must do the spack setup on a netboot node. It's easiest to do on the virtual FE node.

Also I don't mind reinstalling qlustar with the FE VM node but I would appreciate how to run it.

It's automatically started at boot as a systemd service as explained in the first steps doc.

Now here is my second question , how to run the demo VM ?

Here is my frustration sequences of event in case someone want to rewrite that part of the doc ;-)

demo-system -a start
Starting VMs in a tmux session ...
Created tmux session demo-system.
  => Attach with 'console-demo-vms'

So far so good so I ran console-demo-vms

Something happened and got a nice command line in the bottom of the terminal, but it looks like I  am still on the front end , not in the demo VM ,  not only the prompt shows the front end hostname but htop command shows the whole ram of 32GB and CPU #, according to /etc/qlustar/vm-configs/demo.conf

it should be :

CN_MEM=1285
HA_MEM=1024

I am really confuse now, console-demo-vms did not show the demo VM ,

Any reason , what am I doing wrong here ?

console-demo-vms starts a tmux session with each demo node's console in one window. As explained in the first steps doc, you can change to different windows in the tmux session using the <Ctrl-t n> key stroke. That way you can switch to the demo node's consoles.

Now from the head node under root :

0  root@jcluster  ~  #
dsh -a uptime
beo-201: ssh: connect to host beo-201 port 22: No route to host
beo-202: ssh: connect to host beo-202 port 22: No route to host
beo-203: ssh: connect to host beo-203 port 22: No route to host
beo-204: ssh: connect to host beo-204 port 22: No route to host

beo-201 are supposed to be the demo VMs,

The demo nodes have not booted for some reason, you should check their console (see above) for errors.

I have many more questions but that's it for now but planing forward,  I did managed to run QluMan_gui, I played over night (until 5am) trying to add a working user and add a node . etc ..

The doc for me for unusable, there's no tutorial online I could find (other than the first install which is very simple). I would be really appreciate if someone could describe the full procedure to add a node from catching the MAC address of the new node, building an image for that guy and include it in slurm. something to start with .

It's explained in the QluMan docs.

Good luck,

Roland

Thanks for your help .

I really would  love to have qclusar working ASAP for my very urgent project research involving GPU

Thanks

Chris



_______________________________________________
Qlustar General mailing list -- qlustar-general@qlustar.org
To unsubscribe send an email to qlustar-general-leave@qlustar.org