On 6/18/11 16:59 , Dag Blakstad wrote:
I have not [been able to] debug the problem, as the initial error occurs in
a non-deterministic manner (well by current observations, but computers do
not behave non-deteministically :) )
One possibility I thought of is concurrency. I have only run into the
problem on machines with (development environment) with 4 Cores. On single
CPU EC2 it has not occured (yet). The initial problem states "double lock"
and that is what makes me think of this as a potential cause.
That is definitely possible. I thought I had covered all the bases with
concurrency, but it's very hard to know for sure.
What plans do Qi4j team have for persistence regarding eventsourcing? This
issue https://ops4j1.jira.com/browse/QI-280 seems related.
That is the issue for the current EventStore. For persistence, I have a
ton of ideas, but not sure what are good ones. One idea is to go
low-tech, and use something like JDBM only for the working set (if even
that), and then store events in text files (1000 events or so).
Our team is now trying a different design, with a separate entity the data
that was modeled into a Mixin. The problem has so far only occured when this
Mixin was added to an existing Entity, which has several mixins already.
Could there be a problem with size of metadata inferred by the number of
mixins and properties/relations? I will let you know if this
solution/workaround eliminates the error.
Try it out and see what happens.
/Rickard
_______________________________________________
qi4j-dev mailing list
[email protected]
http://lists.ops4j.org/mailman/listinfo/qi4j-dev
