agent-substrate/substrate
E2E test flaky: TestActorDirectAccess/via_ingress
クローズ
#723 opened on 2026/08/04
area/dev-infragood first issuekind/bug
Repository metrics
- Stars
- (1,230 個のスター)
- PR merge metrics
- (平均マージ 3d 13h) (30d で 153 merged PRs)
説明
The E2E test TestActorDirectAccess/via_ingress in internal/e2e/suites/networking/networking_test.go:53 fails intermittently in CI with an ingress connection timeout.
Error
networking_test.go:53: Actor access through ingress returned HTTP 503, want 200; body: upstream connect error or disconnect/reset before headers. reset reason: connection timeout
--- FAIL: TestActorDirectAccess (5.72s)
--- FAIL: TestActorDirectAccess/via_ingress (5.03s)
FAIL github.com/agent-substrate/substrate/internal/e2e/suites/networking
Sample failures
- https://github.com/agent-substrate/substrate/actions/runs/30873771896/job/91880901172 (branch:
docs/pr-etiquette) - https://github.com/agent-substrate/substrate/actions/runs/30824344855 (branch:
ate-router-connect)
Both fail with the same 503/connection timeout against the ingress path. The direct subtest passes in both cases — only via_ingress is affected.
Observed pass rate
Passed in runs 30835856539, 30818259089, 30817878192 — fails roughly 2 out of every 6 observed runs.
Likely cause
The ingress controller or Envoy proxy is not fully ready when the test probes it, causing an intermittent 503. The test likely needs a readiness wait or retry loop before asserting the 200 response via ingress.