FailoverClientConnectionFactory.FailoverTcpConnection.findAConnection() can get caught in infinite loop

[squawk979]

say we have 2 connection factories s1 and s2, initially we connect ok to s1

however a subsequent send results in an error on s1 (eg disconnection), findAConnection is called to establish a new connection

1/ we have lastFactoryToTry = s1, we enter the loop and try nextfactory = s2 but connection to this fails and throws exception. in the exception handling lastFactoryToTry does not equal nextfactory so we reset iterator and loop round.

2/ nextfactory now =s1 connection fails and throws exception. in the exception handling !this.factoryIterator.hasNext() is not true so we loop round. back to 1/

[squawk979]

It's pretty ugly but I think changing:

	if (!this.factoryIterator.hasNext()) {
		if (retried && (lastFactoryToTry == null || lastFactoryToTry.equals(nextFactory))) {

to

	if (retried && ((!this.factoryIterator.hasNext() && lastFactoryToTry == null) || lastFactoryToTry.equals(nextFactory))) {

solves the issue

[artembilan]

@squawk979 ,

Please, raise a PR with a unit test to confirm the problem and with the fix you think makes sense.

Thank you for feedback!

[garyrussell]

Reproduced with this test...

	@Test
	void failoverAllDeadIntegration() throws Exception {
		ServerSocket ss1 = ServerSocketFactory.getDefault().createServerSocket(0);
		ServerSocket ss2 = ServerSocketFactory.getDefault().createServerSocket(0);
		int port2 = ss2.getLocalPort();
		ss2.close();
		ThreadPoolTaskExecutor exec = new ThreadPoolTaskExecutor();
		exec.initialize();
		Future<Boolean> done = exec.submit(() -> {
			Socket accepted = ss1.accept();
			BufferedReader br = new BufferedReader(new InputStreamReader(accepted.getInputStream()));
			br.readLine();
			accepted.getOutputStream().write("ok\r\n".getBytes());
			accepted.close();
			ss1.close();
			return true;
		});
		TcpNetClientConnectionFactory cf1 = new TcpNetClientConnectionFactory("localhost", ss1.getLocalPort());
		TcpNetClientConnectionFactory cf2 = new TcpNetClientConnectionFactory("localhost", port2);
		FailoverClientConnectionFactory fccf = new FailoverClientConnectionFactory(List.of(cf1, cf2));
		CountDownLatch latch = new CountDownLatch(1);
		fccf.registerListener(msf -> {
			latch.countDown();
			return false;
		});
		fccf.start();
		fccf.getConnection().send(new GenericMessage<>("test"));
		assertThat(done.get(10, TimeUnit.SECONDS)).isTrue();
		assertThat(latch.await(10, TimeUnit.SECONDS)).isTrue();
		assertThatExceptionOfType(UncheckedIOException.class).isThrownBy(() ->
		fccf.getConnection().send(new GenericMessage<>("test")));
	}

I'll take this.