Thanks. I am trying to debug an issue when in random intervals under zero load all 16 Notifiers for 16 queues go down with:
** (exit) %DBConnection.ConnectionError{message: "tcp async_recv: closed", severity: :error, reason: :error}
and DBConnection number spikes well beyond pool_size (50) + number of queues (16) to about a 100. Then after 10 minutes goes down to pool_size + number of queues.
Do workers require even more connections? Or will they use the pool ones? I am just trying to understand what is happening.
The weirdness is of course just a subjective feeling. I understand the reason for notifiers requiring their own connection. It’s just that one would expect that pool_size is the max that can happen.
Update: It’s deployed in Azure and the issue might be (missing) connection keepalive.