This was prompted by the recent failure of Pace to restart the fetch loop. I'm still not exactly sure why that happened, but I suspect Redis. In order to sort out the issue more easily next time, I'm doping all the log lines with the "[pace]" tag so they stand out a little more clearly.
Also, this patch attempts to make the initial loop setup more reliable by attaching it to a callback on the EM::Hiredis client.
This was prompted by the recent failure of Pace to restart the fetch loop. I'm still not exactly sure why that happened, but I suspect Redis. In order to sort out the issue more easily next time, I'm doping all the log lines with the "[pace]" tag so they stand out a little more clearly.
Also, this patch attempts to make the initial loop setup more reliable by attaching it to a callback on the EM::Hiredis client.