Mnesia startup woes

[Evans, Matthew]

Hi,

I have the following mnesia deployment. Two nodes (NodeA and NodeB) with disc_copies and disc_only_copies of tables (other nodes have ram copies of these tables too). 

NodeA starts first, followed by NodeB. NodeA is considered by the non-Erlang part of the system to be a master, and NodeB as the standby. When NodeB comes up (maybe much later) it calls the following code to get the schema from NodeA, join the party and copy the tables to itself (it will remove all files from the mnesia directory first):

    mnesia:change_config(extra_db_nodes, [NodeA]),
    mnesia:change_table_copy_type(schema, node(), disc_copies),
    Tabs = mnesia:system_info(tables) -- [schema],
    [mnesia:add_table_copy(Tab, node(), disc_copies) || Tab <- Tabs].

We can not control when NodeB starts, if ever. So it's likely that NodeA will be the only disc_only node when the tables are created.

What I notice is that if NodeA comes up first, the data is unusable until NodeB comes up. This makes the system unusable for that period of time. Certainly I can understand why we do that (we need consistent data), but what I was after is a way to reverse mnesia:change_config/2 to remove NodeB?

I can do this on NodeA:

[ mnesia:force_load_table(T)  || T <- mnesia:system_info (tables) ].

But was wondering if there is a better way?  For example, to remove NodeB from NodeA's schema?

Thanks