From 0f239f49c6d6c38447ec51698701ff06197771de Mon Sep 17 00:00:00 2001
From: embwl0x <embwl0x@users.noreply.github.com>
Date: Tue, 14 Jul 2026 18:09:16 -0500
Subject: [PATCH] fix(gateway): stop systemd retries on fatal config

---
 hermes_cli/gateway.py                         |  3 +++
 tests/hermes_cli/test_gateway_service.py      |  3 +++
 .../test_systemd_optional_directives.py       | 23 +++++++++++++++++++
 3 files changed, 29 insertions(+)

diff --git a/hermes_cli/gateway.py b/hermes_cli/gateway.py
index 7b9817eadd8f..f45a0e85fb43 100644
--- a/hermes_cli/gateway.py
+++ b/hermes_cli/gateway.py
@@ -32,6 +32,7 @@ PROJECT_ROOT = Path(__file__).parent.parent.resolve()
 from gateway.status import terminate_pid
 from gateway.restart import (
     DEFAULT_GATEWAY_RESTART_DRAIN_TIMEOUT,
+    GATEWAY_FATAL_CONFIG_EXIT_CODE,
     GATEWAY_SERVICE_RESTART_EXIT_CODE,
     parse_restart_drain_timeout,
 )
@@ -2761,6 +2762,7 @@ Environment="HERMES_HOME={hermes_home}"
 Restart=always
 RestartSec=5
 RestartForceExitStatus={GATEWAY_SERVICE_RESTART_EXIT_CODE}
+RestartPreventExitStatus={GATEWAY_FATAL_CONFIG_EXIT_CODE}
 KillMode=mixed
 KillSignal=SIGTERM
 ExecReload=/bin/kill -USR1 $MAINPID
@@ -2795,6 +2797,7 @@ Environment="HERMES_HOME={hermes_home}"
 Restart=always
 RestartSec=5
 RestartForceExitStatus={GATEWAY_SERVICE_RESTART_EXIT_CODE}
+RestartPreventExitStatus={GATEWAY_FATAL_CONFIG_EXIT_CODE}
 KillMode=mixed
 KillSignal=SIGTERM
 ExecReload=/bin/kill -USR1 $MAINPID
diff --git a/tests/hermes_cli/test_gateway_service.py b/tests/hermes_cli/test_gateway_service.py
index ef563c34b49a..6dd6a2152314 100644
--- a/tests/hermes_cli/test_gateway_service.py
+++ b/tests/hermes_cli/test_gateway_service.py
@@ -14,6 +14,7 @@ import hermes_cli.gateway as gateway_cli
 from gateway import status
 from gateway.restart import (
     DEFAULT_GATEWAY_RESTART_DRAIN_TIMEOUT,
+    GATEWAY_FATAL_CONFIG_EXIT_CODE,
     GATEWAY_SERVICE_RESTART_EXIT_CODE,
 )
 
@@ -435,6 +436,7 @@ class TestGeneratedSystemdUnits:
         assert "ExecStop=" not in unit
         assert "ExecReload=/bin/kill -USR1 $MAINPID" in unit
         assert f"RestartForceExitStatus={GATEWAY_SERVICE_RESTART_EXIT_CODE}" in unit
+        assert f"RestartPreventExitStatus={GATEWAY_FATAL_CONFIG_EXIT_CODE}" in unit
         # TimeoutStopSec must exceed the default drain_timeout (60s) so
         # systemd doesn't SIGKILL the cgroup before post-interrupt cleanup
         # (tool subprocess kill, adapter disconnect) runs — issue #8202.
@@ -546,6 +548,7 @@ class TestGeneratedSystemdUnits:
         assert "ExecStop=" not in unit
         assert "ExecReload=/bin/kill -USR1 $MAINPID" in unit
         assert f"RestartForceExitStatus={GATEWAY_SERVICE_RESTART_EXIT_CODE}" in unit
+        assert f"RestartPreventExitStatus={GATEWAY_FATAL_CONFIG_EXIT_CODE}" in unit
         # TimeoutStopSec must exceed the default drain_timeout (60s) so
         # systemd doesn't SIGKILL the cgroup before post-interrupt cleanup
         # (tool subprocess kill, adapter disconnect) runs — issue #8202.
diff --git a/tests/hermes_cli/test_systemd_optional_directives.py b/tests/hermes_cli/test_systemd_optional_directives.py
index 34aa1793281c..cbefa4f2e4c6 100644
--- a/tests/hermes_cli/test_systemd_optional_directives.py
+++ b/tests/hermes_cli/test_systemd_optional_directives.py
@@ -139,6 +139,29 @@ WantedBy=default.target
 
 
 class TestSystemdUnitIsCurrent:
+    def test_unit_without_fatal_config_restart_policy_is_not_current(
+        self, tmp_path, monkeypatch,
+    ):
+        from hermes_cli import gateway as gw
+
+        expected = """[Service]
+Restart=always
+RestartForceExitStatus=75
+RestartPreventExitStatus=78
+"""
+        installed = expected.replace("RestartPreventExitStatus=78\n", "")
+        unit_file = tmp_path / "hermes-gateway.service"
+        unit_file.write_text(installed)
+
+        monkeypatch.setattr(gw, "get_systemd_unit_path", lambda system=False: unit_file)
+        monkeypatch.setattr(
+            gw,
+            "generate_systemd_unit",
+            lambda system=False, run_as_user=None: expected,
+        )
+
+        assert gw.systemd_unit_is_current(system=False) is False
+
     def test_unit_without_optional_directives_is_current(self, tmp_path, monkeypatch):
         """Installed unit missing RestartMaxDelaySec/RestartSteps should be
         considered current when the generated unit includes them."""