2017-10-18 23:41:57
good
The horizon service of the Blackhole project is an essential part. Every connection
is going through it. We seldom need to restart it. Configuration changes can be made
dynamically. No restart needed. But sometime, because of a bug or a limitation, we have to
do it. It is not possible to do it gracefully unless the server is removed from the top
level cluster. Further, every containers inject at startups times extra configurations.
Well, the container can't do this (no access), but the script starting the container do it.

We reworked the container startup script to split the horizon configuration into a separate
script file. When restarting the horizon, we simply walk all the containers and replay its
horizon script. While, restarting the horizon is not completely transparent, it is reliable.