[mgaffigan]

Adds two load tests:

ChainFanOut: File Reader -> MLLP and HTTP across three channels

QueuedDestinations: Parallel destination threads

Both calculate metrics on 1k messages.

[github-actions]

Test Results

130 files  ± 0  130 suites  ±0   12m 42s ⏱️ + 9m 40s
791 tests + 2  788 ✅  - 1  0 💤 ±0  0 ❌ ±0  3 🔥 +3 
929 runs  +14  916 ✅ +7  6 💤 ±0  1 ❌ +1  6 🔥 +6 

For more details on these errors, see this check.

Results for commit 4ba7049. ± Comparison against base commit a403039.

♻️ This comment has been updated with latest results.

[NicoPiel]

1. Flaky queue-depth check (QueuedDestinationsLoadTest.java:736): the test only starts looking at queue depth after all messages have been received. On a fast host the queues may already be empty by then, so maxQueueDepth > 0 fails even though the queues did fill up. For the same reason, the drain time it reports is wrong.
2. Cleanup can leave channels running (ChainFanoutLoadTest.java:115 and QueuedDestinationsLoadTest.java:713): clean() runs before the undeploys without a try/finally. If it throws, the load channels stay deployed, with MLLP 9082, HTTP 9083 and a File Reader still active on the shared server.
3. Server change loses error detail (MirthResourceInvocationHandlerProvider.java:251): the new SQLException handling keeps only the message. It drops the SQLState, the vendor code and the getNextException() chain, and it logs at ERROR on every hit. It's also a change to production error handling bundled into a test PR, with no unit test.