AbstractDatabaseManager: a manager whose startup failed still receives writes, and is never shut down

[ramanathan1504]

Description

AbstractDatabaseManager checks isRunning() on the shutdown path but not on the write path.

After startupInternal() throws, startup() logs the cause and leaves running = false. Then:

• write() (line 297) does not check it — keeps accepting events and dereferences state startup never assigned. NPE per event, naming an internal field, burying the one line that named the cause.
• shutdown() (line 230) does check it — so shutdownInternal() never runs and whatever startup built is leaked. Returns true anyway.

Affects every subclass whose startup fails: JDBC, JPA, NoSQL. #4242 is the case where the leak prevents JVM exit.

Version: 2.x 04c93c1d33, and main — unchanged in this respect.

Logs

main ERROR CassandraManager Could not perform database startup operations:
  InvalidQueryException: Keyspace 'log4j' does not exist
    at AbstractDatabaseManager.startup(AbstractDatabaseManager.java:259)

main ERROR Unable to write to database [Cassandra] for appender [Cassandra].
java.lang.NullPointerException: Cannot invoke "PreparedStatement.bind(Object[])"
  because "this.preparedStatement" is null        <- then this, per event

Suggested fix

• write(): return early when !isRunning(), logging once rather than per event.
• shutdown(): run shutdownInternal() regardless of running, since startup may have acquired resources before throwing.