[Planetlab-users] Re: can't login to nodes on slice..again..and again...and..again

[Anirban Banerjee]

Hi friends,

Thanks to all who responded :) . This is just a suggestion .. maybe one for
the PL operations/admin group to deliberate:  can we try to evolve a 'QoS '
reservation mechanism for PL. I realize that I am using the word QoS for the
lack of a better sounding TLA.

I think that as an engineer and researcher, it is imperative that we be able
to come up with  "reproducible" results from our implementations. A small
example: I implemented a basic DHT on PL to hash files to nodes. Now any
results that I observe (n/w latency, CPU time..etc) are good for (say) the
first 2-3 times I run my experiments. The 4th time some nodes/links fail and
boom I have to replace the failed node (automatically with a script) and
re-make the DHT/links/neighbor table etc.. I realize that I could I have
used existing code on PL, but the point is : If I report results from PL,
they should be reproducible. It seems flaky to report that average latency
is x sec, till some crappy node /link screws up the curve ... and so on.

Again, I am not suggesting that these issues can't be handled (remove
outliers, take averages of readings..etc..)  but its inconvinient. Should we
think about some kind of guarantees that users should be getting regarding
resources on PL? There seem to be some individual projects dealing with QoS
on PL, but does anyone know of any systemwide mechanism to reserve b/w, CPU
resource & nodes..and so on?

Cheers everyone,
-A


On 2/24/07, Anirban Banerjee <banerjee.anirban at gmail.com> wrote:
>
> Hi Everyone,
>                    I often wonder if people using PL  find it hard to
> secure relatively large numbers of nodes for their work. I have about 228
> nodes registered for my slice and can log into only 80 odd nodes... ! !
>
> The .ssh directory permissions are 0700, I cleaned out the known_hosts
> file in case that was causing conflicts, I also  added
> StrictHostKeyChecking No to the ssh config file... inspite of this I can't
> login to more than half of the registered nodes.. :( . Iread the wiki, and
> couldn't get any more insight on this issue.
>
> Is there a quick fix solution to this thing? I added all these nodes at
> least 4-5 days ago.. so I expect that any information that trickles down to
> them regarding my keys etc. must be on the nodes by now.
>
> Do most people target 'x' nodes to impement experiments on and then
> register 2x/3x+ nodes on their slices.."hoping for the best" ??
> It'll be good to hear from peers about how they deal with such a situation
> :) .
>
> Thanks for your time,
> -A
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.planet-lab.org/pipermail/users/attachments/20070226/9735527f/attachment.html