feat(worker-rejoin-resync): wire config fields into xceadmin build (task 1.5)

Add workerResyncWindowMs/workerResyncTimeoutMs to xce.schema.json
(minimum:0, matching the optional-scalar-field precedent -- these
aren't `required` with a floor like heartbeatInterval, so minimum:1
would be wrong) and to mapping.py's YAML->capnp field table. Regenerate
both golden config-build fixtures (task 1.1 grew ClusterConfig's wire
layout by 2 fields; full/expected.bin and minimal/expected.bin each
grew by exactly 8 bytes). full's fixture YAML now sets real values
(900000ms/15000ms); minimal deliberately left untouched (no xce.yaml)
per its own locked-down "hosts+queues only" boundary test -- its
expected.bin still regenerates to reflect the new default-0 fields.
Two new tests assert specific decoded values, not just "build succeeds".

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Y4M5MJM2zK6nU43da267Ku

src/configs/schema/xce.schema.json

@@ -29,6 +29,16 @@
       "type": "integer",
       "minimum": 0,
       "description": "network I/O timeout in seconds"
+    },
+    "worker_resync_window_ms": {
+      "type": "integer",
+      "minimum": 0,
+      "description": "worker job-state reconciliation: how far back (in milliseconds) a recently-finished job stays eligible for extra reporting on rejoin"
+    },
+    "worker_resync_timeout_ms": {
+      "type": "integer",
+      "minimum": 0,
+      "description": "worker job-state reconciliation: how long (in milliseconds) the controller waits for a rejoining worker's job-status report before resolving the reconciliation as timed out"
     }
   },
   "required": ["cluster_name", "heartbeat_interval", "inactive_heartbeat_count"]

src/xceadmin/src/xcecommon/mapping.py

@@ -13,7 +13,7 @@ entries -- ``hosts.yaml`` is the one file that feeds two separate capnp
 fields, ``hgroups`` and ``hostConfigs``; see the Data Models table in
 design.md). Two mapper shapes cover every section:
 
-- :func:`map_scalar_section` -- the single flat ``xce.yaml`` section, 5
+- :func:`map_scalar_section` -- the single flat ``xce.yaml`` section, 7
   top-level ``ClusterConfig`` scalar fields.
 - :func:`map_list_section` -- the 7 remaining sections, each a YAML list
   mapped to a capnp ``List(Struct)`` field, preserving YAML source order
@@ -180,6 +180,12 @@ _XCE_SCALAR_FIELDS: tuple[tuple[str, str], ...] = (
     ("inactive_heartbeat_count", "inactiveHeartbeatCount"),
     ("fetch_job_interval", "fetchJobInterval"),
     ("netio_timeout", "netioTimeout"),
+    # worker-rejoin-resync (Requirements 3.7, 5.3): shared, cluster-wide
+    # via operational-uniformity override (design.md Configuration
+    # Model) -- same domain as heartbeat_interval/inactive_heartbeat_count
+    # above, wired the same way (task 1.5).
+    ("worker_resync_window_ms", "workerResyncWindowMs"),
+    ("worker_resync_timeout_ms", "workerResyncTimeoutMs"),
 )
 
 # resources.yaml ``type``/``order`` -> Resource.valtype/ordtype, mirroring
@@ -500,7 +506,7 @@ def map_list_section(
 def map_scalar_section(
     spec: SectionSpec, values: dict | None, target: "capnp._DynamicStructBuilder"  # noqa: ARG001
 ) -> None:
-    """Set each of the 5 ``xce.yaml``-sourced scalar ``ClusterConfig``
+    """Set each of the 7 ``xce.yaml``-sourced scalar ``ClusterConfig``
     fields directly from ``values`` (:data:`_XCE_SCALAR_FIELDS`).
 
     ``spec`` is accepted for interface symmetry with :func:`map_list_section`

src/xceadmin/tests/fixtures/full/configs/xce.yaml

@@ -4,3 +4,5 @@ heartbeat_interval: 30
 inactive_heartbeat_count: 3
 fetch_job_interval: 10
 netio_timeout: 30
+worker_resync_window_ms: 900000
+worker_resync_timeout_ms: 15000

src/xceadmin/tests/test_config_build.py

@@ -83,6 +83,50 @@ def test_build_same_input_twice_is_byte_identical(tmp_path):
     assert first == second
 
 
+# ---------------------------------------------------------------------------
+# worker-rejoin-resync (Requirements 3.7, 5.3): workerResyncWindowMs/
+# workerResyncTimeoutMs are operator-settable through xce.yaml's scalar
+# section, following the heartbeat_interval/inactive_heartbeat_count
+# pattern exactly (task 1.5). These prove the VALUE round-trips
+# (YAML -> binary -> decode) with a specific, checkable pair of numbers --
+# not just that some golden blob happens to match.
+# ---------------------------------------------------------------------------
+
+
+def test_build_maps_worker_resync_window_and_timeout_fields(tmp_path):
+    _write_yaml(
+        tmp_path / "xce.yaml",
+        {
+            "cluster_name": "eda1",
+            "heartbeat_interval": 30,
+            "inactive_heartbeat_count": 3,
+            "worker_resync_window_ms": 900000,
+            "worker_resync_timeout_ms": 15000,
+        },
+    )
+
+    data = build.build_cluster_config(str(tmp_path))
+
+    mod = load_schema()
+    with mod.ClusterConfig.from_bytes(data) as reader:
+        assert reader.workerResyncWindowMs == 900000
+        assert reader.workerResyncTimeoutMs == 15000
+
+
+def test_build_leaves_worker_resync_fields_at_default_when_absent(tmp_path):
+    _write_yaml(
+        tmp_path / "xce.yaml",
+        {"cluster_name": "eda1", "heartbeat_interval": 30, "inactive_heartbeat_count": 3},
+    )
+
+    data = build.build_cluster_config(str(tmp_path))
+
+    mod = load_schema()
+    with mod.ClusterConfig.from_bytes(data) as reader:
+        assert reader.workerResyncWindowMs == 0
+        assert reader.workerResyncTimeoutMs == 0
+
+
 # ---------------------------------------------------------------------------
 # Hard failures propagate (schema violation, reserved name, duplicates)
 # ---------------------------------------------------------------------------

src/xceadmin/tests/test_mapping.py

@@ -60,11 +60,12 @@ def test_section_map_hosts_yaml_feeds_two_distinct_capnp_fields():
     assert yaml_keys == {"hostgroups", "hosts"}
 
 
-def test_section_map_covers_all_13_clusterconfig_fields_or_documents_exclusion():
-    # 5 scalar (xce.yaml) + 7 list capnp_fields covers 12 of the 13
-    # ClusterConfig fields; configGeneration (@12) is the one deliberately
-    # excluded field (Requirement 1.9), documented in the module docstring,
-    # not present in SECTION_MAP at all.
+def test_section_map_covers_all_14_clusterconfig_fields_or_documents_exclusion():
+    # 7 scalar (xce.yaml, including worker-rejoin-resync's workerResyncWindowMs/
+    # workerResyncTimeoutMs, task 1.5) + 7 list capnp_fields covers 14 of the
+    # 15 ClusterConfig fields (@0-@14); configGeneration (@12) is the one
+    # deliberately excluded field (Requirement 1.9), documented in the
+    # module docstring, not present in SECTION_MAP at all.
     list_fields = {s["capnp_field"] for s in mapping.SECTION_MAP if s["kind"] == "list"}
     assert list_fields == {
         "resources",
@@ -83,6 +84,8 @@ def test_section_map_covers_all_13_clusterconfig_fields_or_documents_exclusion()
         "inactiveHeartbeatCount",
         "fetchJobInterval",
         "netioTimeout",
+        "workerResyncWindowMs",
+        "workerResyncTimeoutMs",
     }
 
 
@@ -91,7 +94,7 @@ def test_section_map_covers_all_13_clusterconfig_fields_or_documents_exclusion()
 # ---------------------------------------------------------------------------
 
 
-def test_map_scalar_section_sets_all_5_fields(cluster_config):
+def test_map_scalar_section_sets_all_7_fields(cluster_config):
     spec = mapping.SECTION_MAP[0]
     values = {
         "cluster_name": "eda1",
@@ -99,6 +102,8 @@ def test_map_scalar_section_sets_all_5_fields(cluster_config):
         "inactive_heartbeat_count": 3,
         "fetch_job_interval": 10,
         "netio_timeout": 30,
+        "worker_resync_window_ms": 900000,
+        "worker_resync_timeout_ms": 15000,
     }
     mapping.map_scalar_section(spec, values, cluster_config)
 
@@ -107,6 +112,8 @@ def test_map_scalar_section_sets_all_5_fields(cluster_config):
     assert cluster_config.inactiveHeartbeatCount == 3
     assert cluster_config.fetchJobInterval == 10
     assert cluster_config.netioTimeout == 30
+    assert cluster_config.workerResyncWindowMs == 900000
+    assert cluster_config.workerResyncTimeoutMs == 15000
 
 
 def test_map_scalar_section_leaves_missing_optional_keys_at_default(cluster_config):
@@ -117,6 +124,8 @@ def test_map_scalar_section_leaves_missing_optional_keys_at_default(cluster_conf
     assert cluster_config.clusterName == "eda1"
     assert cluster_config.fetchJobInterval == 0
     assert cluster_config.netioTimeout == 0
+    assert cluster_config.workerResyncWindowMs == 0
+    assert cluster_config.workerResyncTimeoutMs == 0
 
 
 @pytest.mark.parametrize("empty", [None, {}])