[Server-devel] Server Admin Interface
martin.langhoff at gmail.com
Thu Feb 28 15:21:16 EST 2008
On Thu, Feb 28, 2008 at 6:00 PM, John Watlington <wad at laptop.org> wrote:
> Yeah, so one would think. ejabberd is the only barrier to being able
> to do so at this time. We can contract to get an erlang script
> for fixing up the mnesia database if we continue to use ejabberd.
There's a resident Erlang/ejabberd expert at Catalyst. I'll see what
he thinks about the whole thing.
> But once laptops are registered with a school, they are registered to
> the FQDN of the school. If you change the domain name
> you will currently orphan a lot of laptops.
> Longer term, I want to see a datastore architecture that doesn't have
> this limitation. A school server is just an intermediate cache of a
200% in agreement. I'll keep that in mind as I learn about the arch we have.
> Moving to another school (or between schools) thus works just
> fine (with a long latency when accessing data not already cached on a
> particular school's servers), as does operation after graduation from
> a school.
Agreed. There's a trick there in making the 'moving' permanent as
opposed to temporary. In any case, if we come up with a half-decent
distributed storage/repluication strategy, it should work
transparently. It will have to -- in a town with several schools, kids
may be covered at home by a mesh net associated to another school.
> But that is still in the formative stages (and requires countries to
> foot the
> bill for large centralized data centers.)
The big disk in the sky :-) We might be able to bypass that...
> A manual override would work fine for starters. It's pretty
> stable !
> I would prefer to spend the effort improving the reliability of
> the servers
> than working on automatic switchover methods. One easy modification
> planned is to move the core OS and services onto a Flash disk, so
> networking continues to operate even after a disk failure.
Agreed. Stable and simple over fancy, anytime.
More information about the Server-devel