Slide 22
Slide 22 text
Loadshedding
Well-behaving even when the system receives more requests than it can handle.
Action Behavior before Behavior after
Incoming requests Requests queue up, delayed
response, client times out.
Limit the queue, fail fast
for excessive requests*
Liveness probe Timeout, Pod restarted by
Kubernetes
Non-Blocking, Pod
survives
* needs to be configured via http-max-queued-requests