[erlang-questions] auto-syncing mnesia after a network split
Mon Dec 8 22:07:52 CET 2008
Jim McCoy wrote:
> Consistence, availability, partition-tolerance: pick two.
> Simiarly, Claes is incorrect in assuming that you could use the system
> clock for anything other than printing a nice little display on the
> screen to remind the user that they have a meeting scheduled for some
> point in the near future. There is _a lot_ of literature out there on
> distributed transaction fault-tolerance and it gets pretty complex in
> a hurry.
I'm not wrong if instead of choosing Consistence, availability
I choose availability, partition-tolerance
I.e sacrifice Consistence. NTP works sufficiently well IMHO.
If system clocks are synchronized - how big are the chances of
picking the wrong new master. Sufficiently slim I'd say.
> To maintain the current properties of mnesia I believe the only option
I never said that - I suggested the properties were wrongly chosen.
I/We should have sacrificed consistency - 10 years ago.
> would be to add some complexity on the mechanics of the join mechanism
> (e.g. a quorum system like paxos to decide membership and agreement
A lot of the mnesia clusters only have 2 nodes - thus making
quorums a no-goer. typical HA telcoms chassis systems have exactly
two management blades.
More information about the erlang-questions