Google App Engine updated health checks - google-app-engine

Question

I am using Google App Engine Flexible environment (Node.js). Is there any reason why both the liveness and readiness checks are fired 6 times each at every interval second specified? (these are all at the same timestamp)
A GET 200 2 B 2 ms GoogleHC/1.0 /readiness_check GET 200 2 B 2 ms GoogleHC/1.0
A GET 200 2 B 2 ms GoogleHC/1.0 /readiness_check GET 200 2 B 2 ms GoogleHC/1.0
A GET 200 2 B 1 ms GoogleHC/1.0 /readiness_check GET 200 2 B 1 ms GoogleHC/1.0
A GET 200 2 B 1 ms GoogleHC/1.0 /readiness_check GET 200 2 B 1 ms GoogleHC/1.0
A GET 200 2 B 3 ms GoogleHC/1.0 /readiness_check GET 200 2 B 3 ms GoogleHC/1.0
A GET 200 2 B 1 ms GoogleHC/1.0 /readiness_check GET 200 2 B 1 ms GoogleHC/1.0
A GET 200 2 B 2 ms GoogleHC/1.0 /liveness_check GET 200 2 B 2 ms GoogleHC/1.0
A GET 200 2 B 2 ms GoogleHC/1.0 /liveness_check GET 200 2 B 2 ms GoogleHC/1.0
A GET 200 2 B 2 ms GoogleHC/1.0 /liveness_check GET 200 2 B 2 ms GoogleHC/1.0
A GET 200 2 B 2 ms GoogleHC/1.0 /liveness_check GET 200 2 B 2 ms GoogleHC/1.0
A GET 200 2 B 1 ms GoogleHC/1.0 /liveness_check GET 200 2 B 1 ms GoogleHC/1.0
A GET 200 2 B 1 ms GoogleHC/1.0 /liveness_check GET 200 2 B 1 ms GoogleHC/1.0
And is it normal for the readiness checks to continue indefinitely? I would have thought the readiness checks would stop after an instance was deemed "ready." Just doesn't seem necessary to have both readiness and liveness checks continuously hitting my instance when just the liveness checks would seemingly suffice. If anyone knows of a better way to configure this so that it isn't so redundant I would greatly appreciate it. The relevant pieces of my app.yaml can be seen below:
runtime: nodejs
env: flex
readiness_check:
path: '/readiness_check'
check_interval_sec: 20
timeout_sec: 4
failure_threshold: 2
success_threshold: 2
app_start_timeout_sec: 300
liveness_check:
path: '/liveness_check'
check_interval_sec: 30
timeout_sec: 4
failure_threshold: 3
success_threshold: 2
initial_delay_sec: 300
Thanks!

Answer 1

The purpose of readiness checks to continue forever is so that you can, at your own will, remove a VM out of traffic rotation (e.g. you want to reload some cache, etc). Same happens with liveness. It also allows the load balancer to detect problems and remove VM out of rotation (if your app crashed and is restarting for instance) or heal it (if, say, the VM died altogether).
Regarding the number of requests - it depends on the number of instances you deployed (are you running 2 instances ?) and it works with 3 parallel requests per instance (from 3 different health checkers) so that we ensure availability and reliability of the results.
In the grand scheme of things, unless you are too aggressive and implement custom health check handlers that are too expensive, this should be minimal traffic for your VM.
Hope it helps,
Andre