[Planetlab-devel] nodes do not boot if PLC is unreachable ?

[Larry Peterson]

Here's how I'd proceed. PLC, managed from Princeton, will
continue to assume good connectivity and a large user base
that we don't know personally -- we'll require contacting PLC
before reboot.

If you have another deployment that operates under different
circumstances, you're free to modify the system to behave
differently. In other words, not all mgmt authorities need adopt
the exact same rules, as long as they inter-operate and the
general AUP/mission remains the same.

BTW, I'd be inclined to weaken this check in only those cases
where the node provides a valuable service even when it's not
connected to the Internet; e.g., it's a server for an isolated
subnet (e.g., a school server in a remote village).

Larry

On Thu, Jun 11, 2009 at 10:10 AM, Luigi Rizzo<rizzo at iet.unipi.it> wrote:
> On Thu, Jun 11, 2009 at 09:58:54AM -0400, Larry Peterson wrote:
>> This was a design decision. Only PLC knows if the current
>> slices on the node are still valid, or if, perhaps one of the
>> slices has been suspended due to bad behavior (or whatever).
>>
>> Obviously this is problematic for poorly connected sites, so we
>> might reconsider, but the above issue is one that we have to
>> keep in mind.
>
> understood the design principle.
>
> I was wondering if there is any quick workaround to let a node
> come up even if disconnected. Demos are one context, but
> nodes with intermittent connectivity might become of some use
> in the not too far future.
>
> cheers
> luigi
>
>> Larry
>>
>> On Thu, Jun 11, 2009 at 6:18 AM, Luigi Rizzo<rizzo at iet.unipi.it> wrote:
>> > Is there any particular reason why PlanetLab nodes do not
>> > proceed in the boot sequence if they cannot talk to the PLC ?
>> > I am talking of a node that has been already installed
>> > correctly, registered to the PLC, has slices instantiated
>> > and so on, and for some reason needs to be rebooted.
>> >
>> > At the reboot, if the PLC is not reachable, we get the message
>> >
>> > ? ?pl_boot: failed to contact primary server, trying backup XXX
>> > ? ?pl_boot generating new nonce
>> > ? ?pl_boot: fetching script from boot server XXX
>> > ? ?pl_boot: curl request failed with error 7:
>> > ? ?curl: (7) couldn't connect to host
>> >
>> > repeating attempts between primary and backup server.
>> >
>> > I am asking because sometimes it would be convenient
>> > to have a node on a laptop (or in a virtual machine) and run
>> > it in a place where the PLC is not reachable, but slice users
>> > should still be able to log in.
>> >
>> > Would it be hard to fix this? What kind of information is being
>> > fetched that could not be saved and reused from the previous run ?
>> >
>> > ? ? ? ?cheers
>> > ? ? ? ?luigi
>> >
>> > _______________________________________________
>> > Devel mailing list
>> > Devel at lists.planet-lab.org
>> > https://lists.planet-lab.org/mailman/listinfo/devel
>> >
>>
>> _______________________________________________
>> Devel mailing list
>> Devel at lists.planet-lab.org
>> https://lists.planet-lab.org/mailman/listinfo/devel
>
> _______________________________________________
> Devel mailing list
> Devel at lists.planet-lab.org
> https://lists.planet-lab.org/mailman/listinfo/devel
>