Hi Pierre,
We use sequoia with oracle sgbd,
2 applications are linked to sequoia, one uses TABLES and the other uses VIEWS, some of this views are bidirectional (UPDATE statements are available)
we succeed in starting controllers, requesting tables and views configuring a static
shema for the backend with a <DatabaseTable> markup for each table or view of
our schema.
In an another hand i have seen Raidb1 level implies a table parsing. Though the
explanation his duplication needs to acquire locks on table when updates are
requested...
1) Is this interpretation correct ?
Yes. Note that this should also work without a static schema if you ask
views to be fetched in your virtual database configuration file.
2) Do we risk some database inconsistency with this configuration ? (If first
answer is yes, i'm afraid the second answer is yes too :( )
If you are using Sequoia 2.x then yes, views are considered as separate
tables. So if you update the view and try to access the tables, the
proper locking will not be done. In Sequoia 3.x, it is possible to
declare a view definition to tell which tables are accessed by a view so
that Sequoia proceeds to the proper locking. However, it is my
understanding that Sequoia 3.x will not be developed further nor supported.
3) If this configuration is risky, what other choice do we have ?
As long as you always update data the same way (always through the view
or always through the tables), you should be fine. Reads should be
properly isolated with Oracle snapshot isolation whatever way you access
them.
If you are updating concurrently through the view and the tables, then
you will have inconsistencies with Sequoia.
Hope this helps,
Emmanuel
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
