From 1bcdfe8779ee02594816b3f5607502aaea8d49c3 Mon Sep 17 00:00:00 2001 From: weichen Date: Tue, 13 Oct 2026 02:14:30 +0000 Subject: [PATCH] Add real artifact rollback and phase reconciliation tests Why: Stub journals do not prove that a real RDM binary rolls back after a switched fault, a failed build, or a candidate that exits on serve. Evidence: SELF unittest of the three new LifecycleRealTests methods exited 0. Staging leaves the old pid and release in place; the next start clears that journal. Switched and an immediate serve exit both restore the previous real binary. Impact: The crash candidate delegates version and identity to the real binary so the gate still sees a capable artifact. Login and rdm.ping stay NOT_RUN. Refs: tests/integration/lifecycle/test_lifecycle_real.py Co-authored-by: Cursor --- .../lifecycle/test_lifecycle_real.py | 117 ++++++++++++++++++ 1 file changed, 117 insertions(+) diff --git a/tests/integration/lifecycle/test_lifecycle_real.py b/tests/integration/lifecycle/test_lifecycle_real.py index 673928ff..ec465b45 100644 --- a/tests/integration/lifecycle/test_lifecycle_real.py +++ b/tests/integration/lifecycle/test_lifecycle_real.py @@ -3,6 +3,7 @@ 由 make test-lifecycle 调用。未设置 SVC_TEST_REAL 时整组跳过，普通 make test 不因此失败。登录与 rdm.ping 不在这里实现：RDM 没有 user/apikey。 """ +import hashlib import importlib.util import json import os @@ -249,6 +250,122 @@ class LifecycleRealTests(tsr.SvcAutoRecoverTests): if (web / "svc.pid").exists(): self.svc("swweb", "stop") + def file_sha256(self, path): + return hashlib.sha256(Path(path).read_bytes()).hexdigest() + + def journal(self, name): + path = self.runtime(name) / "deploy.journal" + return json.loads(path.read_text()) if path.exists() else None + + def set_source(self, name, source): + path = self.runtime(name) / "svc.env" + lines = [] + for line in path.read_text().splitlines(): + if line.startswith("source="): + lines.append("source=" + source) + else: + lines.append(line) + path.write_text("\n".join(lines) + "\n") + + def ensure_git(self): + subprocess.run(["git", "init", "--initial-branch=main"], cwd=self.root, + env=self.env, check=True, stdout=subprocess.DEVNULL) + subprocess.run(["git", "add", "."], cwd=self.root, env=self.env, check=True, + stdout=subprocess.DEVNULL) + subprocess.run(["git", "commit", "-m", "fixture"], cwd=self.root, env=self.env, + check=True, stdout=subprocess.DEVNULL) + + def deploy_fault(self, name, sha, port, phase): + env = dict(self.env) + env["SVC_TEST_FAULTS"] = "1" + env["SVC_FAULT_AT"] = phase + return self.svc(name, "deploy", "--from", "rdm@" + sha, "--port", str(port), + check=False, env=env) + + def test_phase_interrupt_reconciles_on_real_artifacts(self): + """staging 在停旧前退出，旧进程与指针不动；switched 在指针已切后退出，下次 start 回退。""" + port = self.free_port() + result = self.deploy("ph", self.pin["prev_sha"], port, check=False) + self.assertEqual(result.returncode, 0, result.stdout) + pid = self.pid_of("ph") + first = self.release_id("ph") + digest = self.file_sha256(self.runtime("ph") / "current" / "bin" / "agent-system") + row = self.identity_row("ph") + + staged = self.deploy_fault("ph", self.pin["sha"], port, "staging") + self.assertEqual(staged.returncode, 86, staged.stdout) + self.assertIn("故障注入于 staging", staged.stdout) + self.assertEqual(self.pid_of("ph"), pid) + self.assertTrue(self.alive("ph")) + self.assertEqual(self.release_id("ph"), first) + self.assertEqual(self.journal("ph")["phase"], "staging") + resumed = self.svc("ph", "start") + self.assertIn("未切换", resumed.stdout) + self.assertIsNone(self.journal("ph")) + self.assertEqual(self.pid_of("ph"), pid) + self.assertEqual(self.release_id("ph"), first) + + switched = self.deploy_fault("ph", self.pin["sha"], port, "switched") + self.assertEqual(switched.returncode, 86, switched.stdout) + self.assertIn("故障注入于 switched", switched.stdout) + self.assertIn(self.pin["sha"][:7], self.release_id("ph")) + self.assertFalse(self.alive("ph")) + self.assertEqual(self.journal("ph")["phase"], "switched") + rolled = self.svc("ph", "start") + self.assertIn("已回退上一版本", rolled.stdout) + self.assertEqual(self.release_id("ph"), first) + self.assertEqual( + self.file_sha256(self.runtime("ph") / "current" / "bin" / "agent-system"), + digest) + self.assertEqual(self.identity_row("ph"), row) + self.assertTrue(self.alive("ph")) + self.assertIsNone(self.journal("ph")) + self.svc("ph", "stop") + + def test_build_failure_keeps_real_process(self): + self.ensure_git() + port = self.free_port() + self.seed_established("bld", port) + self.set_source("bld", "local@seed") + pid = self.pid_of("bld") + before = self.file_sha256("/proc/%d/exe" % pid) + (self.root / "Makefile").write_text("build:\n\texit 1\n") + result = self.svc("bld", "deploy", "--from", "local", check=False) + self.assertNotEqual(result.returncode, 0, result.stdout) + self.assertIn("构建失败", result.stdout) + self.assertEqual(self.pid_of("bld"), pid) + self.assertTrue(self.alive("bld")) + self.assertEqual(self.file_sha256("/proc/%d/exe" % pid), before) + self.svc("bld", "stop") + + def test_start_exit_rolls_back_to_real_binary(self): + """serve 立即退出的包装会回退。version 与 identity 仍转给真实制品，门禁才看得到能力。""" + self.ensure_git() + port = self.free_port() + self.seed_established("die", port) + self.set_source("die", "local@seed") + real = self.file_sha256(self.real_bin) + shim = self.root / "crash-shim" + shim.write_text( + "#!/bin/bash\n" + "real=%s\n" + "if [ \"${1:-}\" = serve ]; then\n" + " echo 'crash-shim: serve exits immediately' >&2\n" + " exit 1\n" + "fi\n" + "exec \"$real\" \"$@\"\n" % self.real_bin) + shim.chmod(0o755) + (self.root / "Makefile").write_text( + "build:\n\tmkdir -p bin\n\tcp %s bin/agent-system\n" % shim) + result = self.svc("die", "deploy", "--from", "local", check=False) + self.assertNotEqual(result.returncode, 0, result.stdout) + self.assertIn("已回退上一版本", result.stdout) + current = self.runtime("die") / "current" / "bin" / "agent-system" + self.assertEqual(self.file_sha256(current), real) + self.assertTrue(self.alive("die")) + self.assertIsNone(self.journal("die")) + self.svc("die", "stop") + if __name__ == "__main__": unittest.main()