From c673dc8216eb0719d418f405324617e74fa5e4a2 Mon Sep 17 00:00:00 2001 From: Ryan Lymburner Date: Fri, 19 Sep 2025 11:39:02 -0700 Subject: [PATCH] Update FAQ with: "How do I prevent 503 errors during deployments?" --- docs/faq.md | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/docs/faq.md b/docs/faq.md index 10079703..499e2b6f 100644 --- a/docs/faq.md +++ b/docs/faq.md @@ -20,4 +20,8 @@ AWS Gateway API Controller supports Gateway API CRD bundle versions `v1.1` or gr **How do health checks work in multi-cluster deployments?** -In multi-cluster deployments, when you apply a TargetGroupPolicy to a ServiceExport, the health check configuration is automatically propagated to all target groups across all clusters that participate in the service mesh. This ensures consistent health monitoring behavior regardless of which cluster contains the route resource. \ No newline at end of file +In multi-cluster deployments, when you apply a TargetGroupPolicy to a ServiceExport, the health check configuration is automatically propagated to all target groups across all clusters that participate in the service mesh. This ensures consistent health monitoring behavior regardless of which cluster contains the route resource. + +**How do I prevent 503 errors during deployments?** + +When using AWS Gateway API Controller with EKS, customers may experience 503 errors during deployments due to a timing gap between pod termination and VPC Lattice configuration propagation, which affects the time controller takes to deregister a terminating pod. We recommend setting `terminationGracePeriod` to at least 150 seconds and implementing a preStop hook that has a sleep of 60 seconds (but no more than the `terminationGracePeriod`). For optimal performance, also consider setting `ROUTE_MAX_CONCURRENT_RECONCILES` to 10 which further accelerates the pod deregistration process, regardless of the number of targets. \ No newline at end of file