Erlang leader election - code available?
Wed Mar 17 15:30:52 CET 2004
The leader election module gen_leader contains a bug.
I just returned from a chat at Chalmers with a PhD student
that is looking into verification of the gen_leader. He made
clear to me that the algorithm we used was valid for systems
in which nodes go down, but is not correct for a situation in
which nodes restart (not even if the restart is reasonably slow
compared to the recovery).
He found the bug, I see it as my responsibility to find a solution
for the error (a deadlock and sometimes a partitioning in which
two leaders are elected).
Has anyone observed that in practise? The likelyhood that it
appears is probably not very large, but this is a ticking bomb,
I am looking into the matter. If someone already fixed it, let me
Dr Thomas Arts
Software Engineering and Management
IT-university in Gothenburg
Box 8718, 402 75 Gothenburg, Sweden
Tel +46 31 772 6031
Fax +46 31 772 4899
More information about the erlang-questions