Closed miloszwatroba closed 4 days ago
Should this commit be considered a breaking change and released in a new major version or is this overly cautious?
A quick skim over internal codebases and a GitHub code search doesn't seem to reveal any such usages, so I'd consider it relatively safe.
If it turns out I'm wrong and you've ended up here, please open up an issue if this is a problem for your setup.
Fixes #532
This pull request adds fail-open monitoring to our Load Balancing monitors.
To give context about fail open (ref):
Adding this metric will give better visibility into whether the Load Balancer's fail open routing was used during incidents. The metrics added as part of this pull request are in line with the AWS documentation:
UnhealthyRoutingRequestCount
for ApplicationLoadBalancer withLoadBalancer
andTargetGroup
dimensions (ref)UnhealthyRoutingFlowCount
for NetworkLoadBalancer withLoadBalancer
dimension (ref)These metrics are reported conditionally, only when they have nonzero values. Thus, I added a
FILL(metric, 0)
metric math to correctly represent the values on the dashboards.Tested with NetworkLoadBalancer, as that's the setup I have on my account - the ApplicationLoadBalancer values are based only on the documentation.
Discussion point: I added
metricUnhealthyRoutingCount
to theILoadBalancerMetricFactory
interface, which is exported from the package. This means, that this commit theoretically introduced a breaking change for customers that extended the library with their own classes that implement this interface. How do you approach such changes in this library Should this commit be considered a breaking change and released in a new major version or is this overly cautious?By submitting this pull request, I confirm that my contribution is made under the terms of the Apache-2.0 license