test/integration/mesh.test.ts

@@ -340,3 +340,47 @@ test("a machine keeps what it was given when the mesh says nothing about it", {
   assert.ok((await on("laptop", `grep -q anchor.internal /etc/hosts`)).ok,
     "the mesh's names are gone");
 });
+
+test("a machine that fell behind catches up without being named", { skip, timeout: 900_000 }, async () => {
+  // `status` says which machines are not doing what they were told; something has to act on it.
+  // `push --behind` is that something, and it is a command rather than a timer to begin with —
+  // a scheduler is then a scheduler over this, rather than a second path to the same act.
+
+  // Break one machine, in a way that is fixable: a package that does not exist yet.
+  await must("anchor", `printf %s '{"module":"fixable","version":"1","resources":[` +
+    `{"id":"pkg","type":"package","package":"a-package-that-does-not-exist-yet"},` +
+    `{"id":"note","type":"file","path":"/etc/mesh-fixable","content":"here"}]}' > /tmp/fix.json`);
+  await must("anchor", `docker cp /tmp/fix.json mesh-control:/fix.json`);
+  await mesh("module add /fix.json");
+  await mesh("assign laptop fixable");
+  await mesh("push laptop");
+  await new Promise((r) => setTimeout(r, 8000));
+
+  assert.match(await mesh("status"), /laptop\s+failed/, "the machine did not report a failure");
+  // And the resources that COULD be applied were — one broken thing no longer blocks the rest
+  // (novox/hq 04-ISSUES/011).
+  assert.ok((await on("laptop", `test -f /etc/mesh-fixable`)).ok,
+    "a resource after the failing one was never attempted");
+
+  // Nothing is behind on the other machine, so nothing is pushed to it.
+  const named = await mesh("push --behind");
+  assert.match(named, /laptop/, named);
+  assert.doesNotMatch(named, /sent anchor/, `a machine that was fine was pushed to:\n${named}`);
+
+  // Fix the cause, the way somebody would: the module stops asking for the impossible thing.
+  await must("anchor", `printf %s '{"module":"fixable","version":"1","resources":[` +
+    `{"id":"note","type":"file","path":"/etc/mesh-fixable","content":"here"}]}' > /tmp/fix.json`);
+  await must("anchor", `docker cp /tmp/fix.json mesh-control:/fix.json`);
+  await mesh("module add /fix.json");
+
+  // And nobody names the machine.
+  await mesh("push --behind");
+  await new Promise((r) => setTimeout(r, 8000));
+
+  const after = await mesh("status");
+  assert.doesNotMatch(after, /laptop\s+(failed|refused)/,
+    `the machine did not recover:\n${after}`);
+
+  // With nothing behind, it says so rather than doing nothing quietly.
+  assert.match(await mesh("push --behind"), /every machine is doing what it was told/);
+});