Fault Tolerancy Tests

[NickNick]

We're claiming to be fault tolerant, but we haven't proved it so far. Lets create some tests that handle

Losing connection to the database (database goes down)

• ... due to db-restart
• ... due to signals (test all the 'killing' signals, or at least sigsegv and sigterm)
• ... due to losing the connection because network-layer failure.

What to test for each of these scenarios

• Reconnecting (creating a new connection)
• Old connections must be able to recover and reconnect (the user may not end up with "invalid" connections)
• However, ResultSets must be invalid (and know so).
• Statements prepared before losing connection should still work (this is a MySQL feature, we shouldn't need to recreate statements)
• Stored procedures should still work
• Connection properties, i.e. collation, should still be the same for every connection
• feel free to add more!