You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Putting this up for review now; should work as is if you follow
instructions I added to `./examples/hackernews/README.md`.
There's still some manual configuration changes required to scale out
the followers of the Skip service: creating a corresponding
`skip-ingress-X` service fronting each pod of the statefulset, and
adding the routing logic to the reverse proxy.
The `skip-ingress-X` services can easily be scripted, but using the
current setup changing that routing logic requires a HAProxy restart.
I'm working on a more dynamic config for haproxy using maps, but that's
still a WIP.
# Self-register both the control and event streaming server with the haproxy load balancer.
57
+
# Calling 'set server' after 'add server' is redundant on initial scale-up, but necessary for subsequent scale-ups when a server of that name already exists.
58
+
# Enabling HAProxy TCP health checks ensures that servers are taken out of rotation when the system scales down or instances crash/disconnect for other reasons.
59
+
echo "\
60
+
add server skip_control/follower$id $ip:8081 check ;\
61
+
set server skip_control/follower$id addr $ip port 8081 ;\
62
+
enable server skip_control/follower$id ;\
63
+
enable health skip_control/follower$id ;\
64
+
add server skip_stream/follower$id $ip:8080 check ;\
65
+
set server skip_stream/follower$id addr $ip port 8080 ;\
0 commit comments