Timeouts occuring due to inactivity cause a loss of logs (a broken pipe occurs on the next request, unfortunately theres no notice for the current request, and its just lost).
Adding the keep alive for 3 minutes (from what we've seen) keeps the connection alive (or reconnects) if it fails. In our tests this prevents the loss of log entries (and is a relatively small change).
Timeouts occuring due to inactivity cause a loss of logs (a broken pipe occurs on the next request, unfortunately theres no notice for the current request, and its just lost).
Adding the keep alive for 3 minutes (from what we've seen) keeps the connection alive (or reconnects) if it fails. In our tests this prevents the loss of log entries (and is a relatively small change).