[erlang-questions] how: safe recovery from inconsistent_database/running_partitioned_network error

[Igor Goryachev]

Hello everyone,

I am thinking on implementing a safe recovery routine against
inconsistent_database/running_partitioned_network errors which occur
sometimes in our network environment (because of network failures) with
our cluster of 4 nodes powered by erlang/otp. For now working (but
dirty!) solution looks like restarting all except one nodes in order
they could sync mnesia once again. I am thinking of something that
doesn't require to restart nodes or do mnesia:stop/start. The table I
want to resync safely is stored in RAM and is cleaned of foreign nodes
records (after inconsistent database error detected).

Any suggestions?


-- 
Igor Goryachev
Yandex development team.