[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MirageOS-devel] irmin + mirage-vnetif + alcotest + good ol' arp

On 04/09/2015 07:30 PM, Thomas Gazagnaire wrote:
I'm not that far into doing anything interesting with this pile of code, but I 
wanted to solicit your thoughts and feedback given all the mumbling and general 
noise-making about testing we've been doing.

Thanks for your consideration and support. :)
That's awesome!

I haven't checked the code yet, but is the mapping bi-directionnal, i.e. can 
you modify the ARP table in the Git repository, commit and then it's picked up 
by the ARP code?
Yes, it should work this way (once there's more code in place).
If yes, a good way to stress test the system is to (1) generate a big history 
of stuff and (2) use git reset --hard to revert to arbitrary points in the past 
(which are supposed to be consistent) and check that nothing breaks badly (this 
is a good test for intermediate caches for instances). In a distributed 
settings, if you keep track of the state of multiple unikernels, you can try to 
verify that global invariants (involving multiple unikernels) still hold.
Ah, that's a good suggestion.  Thank you!

Not related to testing, but that would be cool to find a clever use for git 
pull/push between unikernels (replacing ARP broadcast maybe?)
I'd been thinking about (or maybe remembered talking about with someone else?) pre-seeding newly booted unikernels with caches reflecting the current state of the local network, which is a less general case of what you propose.


MirageOS-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.