From 39a7a597434ae5e1973182570d170fbb8bb3923f Mon Sep 17 00:00:00 2001 From: David Fridrich Date: Tue, 4 Aug 2026 16:43:11 +0200 Subject: [PATCH] retry flaky e2e specs once The slowest healthy spec in s2i/knative takes 498s against a 600s default Eventually timeout, only 17% headroom, while ginkgo runs 4 specs in parallel on a 4-core runner. Under contention a normally-8-minute spec overruns and the whole job fails, blocking dependabot auto-merge. The imagePullSecrets spec timed out at 600s on three unrelated branches; PR 176 failed it in the merge queue and passed on retry with identical code, so it is a flake and not a regression. --flake-attempts=2 retries a failed spec once and only costs time when a spec actually fails. Left off test-e2e-bundle, which has no failures in the last 93 runs. --- Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Makefile b/Makefile index 5b32126..7c92db9 100644 --- a/Makefile +++ b/Makefile @@ -122,7 +122,7 @@ test: manifests generate fmt vet setup-envtest ## Run tests. .PHONY: test-e2e ## Run e2e tests. test-e2e: ginkgo - $(GINKGO) -v --timeout=1h --label-filter="!bundle" --fail-fast -p ./test/e2e/ + $(GINKGO) -v --timeout=1h --label-filter="!bundle" --fail-fast --flake-attempts=2 -p ./test/e2e/ .PHONY: test-e2e-bundle ## Run bundle e2e tests. test-e2e-bundle: operator-sdk docker-build docker-push bundle bundle-build bundle-push install-olm-in-cluster ginkgo