improving robustness against corrupted tdb graphs

[Hasan Hasan <ha...@trialox.org>]

Dear All

We newly had again a problem with corrupted
urn:x-localinstance:/graph-access.graph which leads to inability to
activate org.apache.clerezza.platform.content.fsadaptor.BundleFsLoader
The impact is that web-resources under CLEREZZA-INF cannot be served
anymore. I don't know why the graph-access.graph gets corrupted so easily.
Have you ever experienced the same?

I wonder if there is a good mechanism to deal with such problem of
corrupted tdb graphs.
I think e.g. that certain graphs upon framework restart may just be
recreated, thus if a graph is corrupted it will be automatically recreated.
For this to work, clerezza has to catch FileException thrown by tdb and
react accordingly.
Maybe we can think of a policy that can be set when creating an MGraph
which allows to define clerezza behavior in such situation.

In any case I think it is necessary to improve the robustness of clerezza
against corruption of graphs.
What do you think?

Regards
Hasan