From fc084c7019939efc2ee60658e4f16c91187b7c21 Mon Sep 17 00:00:00 2001 From: John Coffey Date: Fri, 20 Nov 2026 15:22:39 -0700 Subject: [PATCH 1/2] GitOps test: fail on any failed webmail request Webmail 27.1-rc.11 drains its connections before it stops (K-W5), so an upgrade under open tabs should lose no request at all. Until now the test only reported failed requests, because rc.10 lost one to three on a restart. --- README.md | 2 +- lab/gitops/gitops-test.sh | 6 +++++- 2 files changed, 6 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 4b5cbef..8c68bf9 100644 --- a/README.md +++ b/README.md @@ -53,7 +53,7 @@ Argo CD or Flux instead of Helm, from a Helm repository in the cluster (`lab/gitops/gitops-test.sh`). Nothing is applied again unchanged and no Secret changes across refreshes. The controller puts a hand edit back. An upgrade with every image pinned by digest rolls everything without signing -out webmail tabs left open across it. Then the 27.1 crossing from a 2026.x +out webmail tabs left open across it or failing one of their requests. Then the 27.1 crossing from a 2026.x store, as changes to the controller's spec: refused, then acknowledged. Under Argo CD the chart's generated secrets are off and the Secrets are made first; Flux runs Helm, so the chart keeps its own. CI runs both diff --git a/lab/gitops/gitops-test.sh b/lab/gitops/gitops-test.sh index b7a91cf..12de4a7 100755 --- a/lab/gitops/gitops-test.sh +++ b/lab/gitops/gitops-test.sh @@ -11,7 +11,8 @@ # 2. self-heal: the webmail scaled down by hand comes back to two # 3. an upgrade, 27.1.0-lab.1 to -lab.2 with every image pinned by digest # as a chart release pins them: everything rolls, the upgrade hook -# runs, and webmail tabs open across it are never signed out +# runs, and webmail tabs open across it are never signed out and +# no request of theirs fails # 4. after it: the same Secrets, mail works, and steady again # # lab/gitops/gitops-test.sh TOOL @@ -154,6 +155,9 @@ for e in r["errors"][:10]: if r["signins"] != 1: print("!! the upgrade signed the tabs out", file=sys.stderr) sys.exit(1) +if r["failed"]: + print("!! a webmail request failed during the upgrade", file=sys.stderr) + sys.exit(1) PY echo "== 4. after the upgrade" -- 2.54.0 From 92a636e3e8f568389679fbfe25e1f660c98b3607 Mon Sep 17 00:00:00 2001 From: John Coffey Date: Fri, 20 Nov 2026 15:54:05 -0700 Subject: [PATCH 2/2] fresh-store: a port-forward that loses the race for 18080 is started again The setup pod's port-forward started as soon as the bootstrap pod's was killed, on the same port. When the old one hadn't let go of 18080 yet, the new one exited and nothing answered for 60 s, though the pod was up: the crossing failed that way under Argo CD on amd64. stop_forward now waits for the port-forward to exit, and forward() starts it again if it has. --- lab/chart/fresh-store.sh | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) diff --git a/lab/chart/fresh-store.sh b/lab/chart/fresh-store.sh index e6d3732..ba42b1e 100755 --- a/lab/chart/fresh-store.sh +++ b/lab/chart/fresh-store.sh @@ -39,19 +39,28 @@ run_pod() { "${K[@]}" -n inbuxa wait --for=condition=Ready "pod/$name" --timeout=180s >/dev/null } -# Port-forward 18080 to a pod's 8080 until stop_forward; waits for it to answer +# Port-forward 18080 to a pod's 8080 until stop_forward; waits for it to +# answer. The port-forward is started again if it exits, as it does when +# the one before it hasn't let go of 18080 yet. FORWARD_PID="" forward() { - "${K[@]}" -n inbuxa port-forward "pod/$1" 18080:8080 >/dev/null 2>&1 & - FORWARD_PID=$! for _ in $(seq 1 60); do + if [ -z "$FORWARD_PID" ] || ! kill -0 "$FORWARD_PID" 2>/dev/null; then + "${K[@]}" -n inbuxa port-forward "pod/$1" 18080:8080 >/dev/null 2>&1 & + FORWARD_PID=$! + fi curl -fsS -o /dev/null http://127.0.0.1:18080/healthz/live 2>/dev/null && return 0 sleep 1 done echo "!! $1 never answered on 8080" >&2 return 1 } -stop_forward() { [ -n "$FORWARD_PID" ] && kill "$FORWARD_PID" 2>/dev/null; FORWARD_PID=""; } +stop_forward() { + [ -n "$FORWARD_PID" ] || return 0 + kill "$FORWARD_PID" 2>/dev/null + wait "$FORWARD_PID" 2>/dev/null || true + FORWARD_PID="" +} trap stop_forward EXIT # The one-off pods' config.json (the chart's own pods have theirs) -- 2.54.0