We have a test which failed constantly.

Steps are:

1. Create a durable queue

java -jar /var/dtests/node_data/clients/amqx.jar  queue --host 127.0.0.1:1099 --username admin --password admin --action add --name 'test_queue' --durable

2. Send 20 messages to the queue with ttl 10000 ms:

/var/dtests/node_data/clients/aac0_sender  --timeout 3 --log-msgs dict --broker admin:admin@127.0.0.1:5672 --address 'test_queue' --count 20 --msg-durable True --msg-ttl 10000

3. Check messages in the test queue and in the Expire queue

Expected result: there is 20 messages in the test queue and 0 messages in the Expire queue

Actual result: there is 20 messages in the test queue and 0 messages in the Expire queue

4. Restart broker

5. Check messages in the test queue and in the Expire queue

Expected result: there is 20 messages in the test queue and 0 messages in the Expire queue

Actual result: there is 20 messages in the test queue and 0 messages in the Expire queue

6. Wait for expiring ttl messages (10000 ms + another 30 seconds more)

7. Check messages in the test queue and in the Expire queue

Expected result: there is 0 messages in the test queue and 20 messages in the Expire queue

Actual result: there is 20 messages in the test queue and 0 messages in the Expire queue

I tried to increase the waiting time to 120 seconds or I added the additional consume messages by a client from the test queue, but it didn't help. However sometimes the test started to be passed.

Also I tried to check it manually and I couldn't reproduce it (maybe I'm too slowly or smth extra happens in this case)

Please, could you check if some code related to ttl messages was changed and is it a real bug or not. I tried to run this test on 7.6.0 release and it 100% green.

Also, restarting broker seems an option step, because we have some failures in similar tests where a broker is killed.

Tested on AMQ 7.7.0 CR1 build (rhel 7x)