e2e: add cmd probe recovery test (gost#837)

Verifies dead->alive->dead node transition: a node excluded by a failing cmd probe resumes carrying traffic once the probe flips healthy, without restart.
2026-10-08 12:05:46 +00:00 · 2026-08-01 23:23:28 +08:00
parent f25148f2e0
commit c7d793c619
2 changed files with 96 additions and 0 deletions
@@ -0,0 +1,52 @@
+services:
+- name: proxy
+  addr: :8080
+  handler:
+    type: http
+    chain: my-chain
+  listener:
+    type: tcp
+
+# Two live relays — both forward to the echo server. Relays never go down;
+# liveness is driven solely by the cmd probes reading flag files in the
+# container, so the test can flip node health without restarting anything.
+- name: relay-a
+  addr: 127.0.0.1:18081
+  handler:
+    type: http
+  listener:
+    type: tcp
+
+- name: relay-b
+  addr: 127.0.0.1:18082
+  handler:
+    type: http
+  listener:
+    type: tcp
+
+chains:
+- name: my-chain
+  hops:
+  - name: hop-1
+    selector:
+      strategy: lowestlatency
+      maxFails: 1
+      failTimeout: 10s
+    nodes:
+    - name: node-a
+      addr: 127.0.0.1:18081
+      connector:
+        type: http
+      probe:
+        # exits 1 while /tmp/a_down exists → node marked dead by probe.
+        type: cmd
+        command: "test -f /tmp/a_down && exit 1 || exit 0"
+        interval: 2s
+    - name: node-b
+      addr: 127.0.0.1:18082
+      connector:
+        type: http
+      probe:
+        type: cmd
+        command: "test -f /tmp/b_down && exit 1 || exit 0"
+        interval: 2s