[PL #8320] Re: planetlab in a box

[Aaron Klingaman via RT]

Email Recipients (see http://www.planet-lab.org/Support)
       Requestor: alk at absarokasoft.com
       Ticket Ccs: danny.bickson at gmail.com, jaffe at cs.huji.ac.il, llp at cs.princeton.edu, smuir at cs.princeton.edu

==================================================

Wed Nov 30 14:21:51 2005: Request 8320 was acted upon.
Transaction: Ticket created by alk

Subject: Re: planetlab in a box

Danny Bickson wrote:
> In a nutshell, we neglected the FC2 option due to network driver bug and
> installed a working FC4 server. 
> 
> The most valuable help we can get from you is if you can login to one of our
> node which is problematic and sniff around to tell us what is missing in
> terms of planetlab software/services.
> 
> We think that the slices that are created on server are not exported to the
> nodes. As we assume, this is related to some planetlab software which is
> either missing or not configured probably.
> We managed to solve all the other hardware problem of the blades and now we
> are stuck since we don't know you system good enough. Our installation was
> done from scratch using a clean FC4 and the top of the CVS tree.
> 
> To be more sepcific, the last input I got from elliot is that  pl_mom tries
> to call "curl  'http://127.0.0.1:3100/slicedata'"
> There is no one listening on that port. We got feedback from Aaron that it
> is related to costat but we don't know what that means and how to fix it.

pl_mom is a non-critical service, you are probably fine with ignoring 
this for now.

> 
> Also we cannot find any service who is supposed to download slices-0.5.xml. 
> 

Mostly for Steve, here is what I found:

[root at planet9 log]# ls /vservers/
lost+found  pl_netflow  vserver-reference

in /var/log/pl_nm:
29Nov05 17:22:57.07     initialising rspecs
29Nov05 17:22:57.11 GFY : 
"[/usr/local/planetlab/bin/sidewinder/nm.py:590,?] 'Server(log, 
config).run()' -> 
[/usr/local/planetlab/bin/sidewinder/NmServerBase.py:56,run] 's
elf.timeout_callback(now)' -> 
[/usr/local/planetlab/bin/sidewinder/nm.py:487,timeout_callback] 'if 
t.timeout(now):' -> [/usr/local/planetlab/bin/sidewinder/VServerVmm.py:9
9,__boot_callback] 'if self.boot_slices.items != []:' -> 
[exceptions.AttributeError] 'list' object has no attribute 'items'"

This may be because the PlanetLabConf entries used by pl_box are out of 
date with the live ones. How to synchronize PlanetLabConf entries with a 
private setup hasn't been figured out yet, and a workaround was put in 
its place (see the *.sql file in the pl_box) setup. I can correct that, 
but it might not be the correct fix for the slices problem.

slices-0.5 looks fine on the server, too.