Connectivity
Changed
- Default
max-worker-connections to 0, making it same as max-worker-open-files i.e. max open files (system's limit) / worker-processes - 1024.
This optimizes for high load conditions where it improves performance at the cost of increasing RAM utilization (even on idle). - HorizontalPodAutoscaler was tuned to use
targetMemoryUtilizationPercentage of 80 due to increased RAM utilization with new default for max-worker-connections of 0. - Removed use of
enable-dynamic-certificates CLI flag, it has been deprecated since ingress-nginx 0.26.0 via ingress-nginx PR #4356 - Changed default
error-log-level from error to notice. - Added a link to the README in the sources of Chart.yaml
Changed
- Align graceful termination configuration with changes made in upstream ingress-nginx 0.26.0 (see related PR #4487 and important section in 0.26.0 release notes).
- Make NGINX IC Deployment’s
terminationGracePeriodSeconds configurable and align its default with configmap.worker-shutdown-timeout - Make NGINX IC controller container lifecycle hooks configurable, and change from
sleep 60 to using /wait-shutdown as preStop hook.
- Make
controller.minReadySeconds configurable.
Changed
- Change deployment to use release revision not time for Helm 3 support.
Updated
- Enable Giant Swarm monitoring
- Set default
ingressClass to kong-app. See #48
Changed
- Fix small cluster profile resource requests. (#42)
Changed
- Use cluster.kubernetes.clusterDomain instead of cluster.kubernetes.domain for custom DNS suffix.
Changed
- Set
autopath variable to disabled by default in values file.
Changed
- Disable HPA and PDB for xs clusters since NGINX Deployment resource requests are not set there. (#40)
Changed
- Adjust resource requests, HPA and PDB depending on determined cluster profile; supported cluster profiles include xxs, xs, small, and larger than small or unknown. (#38)
By default, for nginx on:
- xxs clusters - clear resource requests, HPA and PDB are disabled
- xs clusters - clear resource requests, enabled HPA and PDB
- small clusters - have some resource requests, HPA and PDB are enabled
- clusters larger than small or unknown - have decent resource requests i.e. capacity out-of-the-box, and HPA and PDB are enabled.
Changed
- Reintroduced config properties which should have been just deprecated but got dropped prematurely in v1.4.0 (#36)
configmap.annotations-prefixconfigmap.default-ssl-certificateconfigmap.hpa-enabledconfigmap.hpa-max-replicasconfigmap.hpa-min-replicasconfigmap.hpa-target-cpu-utilization-percentageconfigmap.hpa-target-memory-utilization-percentageconfigmap.ingress-class