[erlang-questions] tcerl memory usage, plans for hash-type db

[Ulf Wiger (TN/EAB)]

Joel Reymont skrev:
> On Aug 22, 2008, at 8:57 PM, Scott Lystig Fritchie wrote:
> 
>> Even if TC's disk structures were 100% crash-safe, aren't there still
>> sync problems between the local txn manager's log and any mnesia_ext
>> backend (as the callbacks exist today)?
> 
> 
> Yes!
> 
> Mnesia assumes that everything is fine by the time you get to the  
> backend module. If the backend module crashes then Mnesia is not gonna  
> know about it.

In the rdbms implementation of external_copies, the general
idea was that if something goes wrong in the backend module,
mnesia dumps core.

As far as I can tell, you're right about how tcerl works.


> There's no way of fixing this short of further serious hacking 
 > of the Mnesia proper.

...or dumping mnesia core at errors. This will happen if the
backend module raises an exception.

This is arguably better than having (possibly serious) errors
go undetected.

BR,
Ulf W