Redundancy at Java application level

Question

Dear All,
Just wanted to know that how can we implement redundancy at Java application level in case of one database gets failed?
Please let me know for the same
Waiting for the reply
Regards
Jignesh

Answer 1

If you have a read-only database (you only do selects from it), simply retry the operation in another database if the first one fails. You could do a minimal select first to check if the db responds. E.g. in Oracle the idiom is "select sysdate from dual", in Postgres you can do "select 1", etc. But the server could fail between the test select and the real select, so you need to retry anyway, so the test select is redundant.
Perhaps retry, sleep a while, retry again, sleep, and after a few rounds give up. This gets around short network breaks, but doesn't busy loop forever if the program has a bug and the SQL statement is broken.
If you have a database where you also do inserts and updates, a backup database becomes ...problematic. If one server fails, and you start updating the other, you now have two conflicting data sets. Perhaps this could work if you have some kind of an append-only database - you add rows, like some kind of event logging; then you can merge the two data sets later. This gets somewhat specialized.
In the general case redundancy needs to be implemented as clustering at the db server level, not at the application level.