Validate new dump can be restored before trying

[gregcorbett]

Before the failover pipeline can restore the database from the dump file, it has to drop the existing data.

• For ORACLE, this happens all at once in dropGocdbUser2.sh
• For MariaDB, this happens table by table i.e.:

DROP TABLE IF EXISTS `countries`;
CREATE TABLE `countries` (...);
INSERT INTO `countries` VALUES (...);

This leads to problems (atleast under ORACLE) where if the dump can't be restored, the existing data gets dumped anyway.

I can imagine a similar problem under MariaDB, where by a CREATE TABLE command could fail, which would mean the database would be missing a table and its data.

Before trying to restore the dump to the live database, the failover scripts should try and restore the dump to some dummy dataabse and if (and only if) that's successful, then restore the dump to the live database.

[gregcorbett]

In the MariaDB case, making the import atomic feels like it would be sufficent.

[tofu-rocketry]

But also, keep the last known good dump for recovery purposes.

[gregcorbett]

I think that behaviour was originally intended - but the ORACLE version may have decayed.

I believe the MariaDB version will only replace the archived dump if importing the new dump doesn't cause an exception.