[Bug]: Fluvio client does not detect broken connection while consuming after machine suspend and resume

[ajhunyady]

I started a client instance to listen on a topic/partition

fluvio consume fluvio-stars -k -B

[2021-08-26 19:12:15] 600 [2021-08-26 19:14:17] 600 [2021-08-26 19:16:18] 600 [2021-08-26 19:18:20] 600 [2021-08-26 19:20:22] 600 [2021-08-26 19:22:24] 600 [2021-08-26 19:24:25] 600 [2021-08-26 19:26:27] 600 [2021-08-26 19:28:29] 600

After a while I closed the laptop. When I returned, the connection looked alive... but soon realized that it does not receive traffic.

Expected behavior

(the minimum)

• Exit out of the loop of connection error and print the error: Connection Error

(better)

• Try to reconnect, and give up after a certain number of retries.

[tjtelan]

Caused by #770

[tjtelan]

This should be approached as an automated test.

Start consume, simulate a network failure (e.g. delete the spu pod, and wait for pod to redeploy), verify reconnection and resuming consuming

[nacardin]

I'll check behavior on Mac laptop on 0.9.13

[nacardin]

On MacOS, this behavior is still present as of 0.9.13. Closing the laptop for 1 minutes is fine, consume is still alive. If I close the laptop for about 10 minutes, the consume command is still running, but is not showing new records.

[nacardin]

Let's check netstat to see the state of the socket during this behavior

[morenol]

After suspend and resume in linux:

$ sudo netstat  -ntpe4     | grep 900
tcp        0      0 192.168.1.105:57768     3.211.184.49:9003       ESTABLISHED 1000       146968     15068/fluvio        
tcp        0      0 192.168.1.105:38622     3.211.184.49:9005       ESTABLISHED 1000       144072     15068/fluvio        

[sehz]

defering

[github-actions[bot]]

Stale issue message