Merge remote-tracking branch 'origin/main' into codex/p0-obs-002-20260715

This commit is contained in:
ogt
2026-07-15 00:50:15 +08:00
8 changed files with 250 additions and 13 deletions

View File

@@ -8,6 +8,7 @@ the database boundary.
from __future__ import annotations
import asyncio
import hashlib
import json
import re
@@ -29,6 +30,7 @@ RECONCILIATION_SUMMARY_OPERATION = "asset_capability_reconciliation_completed"
RECONCILIATION_WORK_ITEM_OPERATION = "asset_capability_reconciliation_work_item_created"
FRESHNESS_SLO_SECONDS = 2 * 60 * 60
RECONCILIATION_FRESHNESS_SLO_SECONDS = 36 * 60 * 60
LIVE_READBACK_TIMEOUT_SECONDS = 8.0
STAGE_IDS = (
"source_truth",
@@ -1044,6 +1046,7 @@ async def _load_live_rows(
"created_at": item.get("coverage_created_at"),
}
asset_ids = list(assets_by_id) or [0]
compliance_result = await db.execute(
text(
"""
@@ -1054,7 +1057,7 @@ async def _load_live_rows(
ORDER BY asset_id, dimension, detected_at DESC
"""
),
{"asset_ids": list(assets_by_id) or [0]},
{"asset_ids": asset_ids},
)
for row in compliance_result.mappings().all():
item = dict(row)
@@ -1072,9 +1075,11 @@ async def _load_live_rows(
FROM asset_change_event
WHERE detected_at >= NOW() - INTERVAL '36 hours'
AND asset_id IS NOT NULL
AND asset_id = ANY(:asset_ids)
ORDER BY detected_at DESC
"""
)
),
{"asset_ids": asset_ids},
)
for row in changes_result.mappings().all():
item = dict(row)
@@ -1112,6 +1117,7 @@ async def build_asset_capability_matrix_with_live_readback(
*,
project_id: str = DEFAULT_PROJECT_ID,
repo_root: Path | None = None,
timeout_seconds: float = LIVE_READBACK_TIMEOUT_SECONDS,
) -> dict[str, Any]:
"""Build the matrix from live DB truth, failing closed to declared scope."""
@@ -1120,12 +1126,17 @@ async def build_asset_capability_matrix_with_live_readback(
)
scope_classes = get_ai_automation_program_scope_classes()
bounded_timeout_seconds = max(0.001, float(timeout_seconds))
try:
latest_run, live_assets, receipt = await _load_live_rows(project_id)
latest_run, live_assets, receipt = await asyncio.wait_for(
_load_live_rows(project_id),
timeout=bounded_timeout_seconds,
)
except Exception as exc:
logger.warning(
"asset_capability_matrix_live_readback_degraded",
error_type=type(exc).__name__,
timeout_seconds=bounded_timeout_seconds,
)
payload = build_asset_capability_matrix(
scope_classes,
@@ -1133,8 +1144,9 @@ async def build_asset_capability_matrix_with_live_readback(
live_readback_status="degraded",
)
payload["live_readback_error_type"] = type(exc).__name__
payload["live_readback_timeout_seconds"] = bounded_timeout_seconds
return payload
return build_asset_capability_matrix(
payload = build_asset_capability_matrix(
scope_classes,
live_assets=live_assets,
latest_run=latest_run,
@@ -1142,3 +1154,5 @@ async def build_asset_capability_matrix_with_live_readback(
repo_root=repo_root,
live_readback_status="ready",
)
payload["live_readback_timeout_seconds"] = bounded_timeout_seconds
return payload

View File

@@ -1,8 +1,11 @@
from __future__ import annotations
import asyncio
from datetime import UTC, datetime, timedelta
from pathlib import Path
from typing import Any
import src.services.ai_automation_asset_capability_matrix as matrix_service
from src.jobs.asset_capability_reconciliation_job import (
build_reconciliation_summary_payload,
build_reconciliation_work_item_input,
@@ -10,7 +13,9 @@ from src.jobs.asset_capability_reconciliation_job import (
from src.services.ai_automation_asset_capability_matrix import (
REQUIRED_CONTROL_IDS,
STAGE_IDS,
_load_live_rows,
build_asset_capability_matrix,
build_asset_capability_matrix_with_live_readback,
build_reconciliation_candidates,
)
from src.services.awoooi_priority_work_order_readback import (
@@ -279,3 +284,132 @@ def test_priority_overlay_projects_aia_p0_006_runtime_controls() -> None:
assert work_item["runtime_progress"]["required_control_count"] == len(
REQUIRED_CONTROL_IDS
)
class _FakeMappings:
def __init__(self, rows: list[dict[str, Any]]) -> None:
self._rows = rows
def first(self) -> dict[str, Any] | None:
return self._rows[0] if self._rows else None
def all(self) -> list[dict[str, Any]]:
return self._rows
class _FakeResult:
def __init__(self, rows: list[dict[str, Any]]) -> None:
self._rows = rows
def mappings(self) -> _FakeMappings:
return _FakeMappings(self._rows)
class _FakeLiveReadbackDb:
def __init__(self, now: datetime) -> None:
self.now = now
self.calls: list[tuple[str, dict[str, Any]]] = []
async def execute(
self,
statement: Any,
params: dict[str, Any] | None = None,
) -> _FakeResult:
sql = str(statement)
bound = params or {}
self.calls.append((sql, bound))
if "FROM asset_discovery_run" in sql:
return _FakeResult(
[
{
"run_id": "00000000-0000-0000-0000-000000000001",
"status": "success",
"ended_at": self.now,
}
]
)
if "FROM asset_inventory" in sql:
return _FakeResult(
[
{
"asset_id": 42,
"asset_key": "host:99",
"asset_type": "host",
"environment": "prod",
"namespace": "",
"name": "99",
"owner_team": "platform_sre",
"source_repo": "wooo/awoooi",
"source_commit_sha": "abc123",
"lifecycle_state": "active",
"last_seen_at": self.now,
"dimension": "auto_monitoring",
"coverage_status": "green",
"coverage_created_at": self.now,
}
]
)
if "FROM asset_compliance_snapshot" in sql:
return _FakeResult([])
if "FROM asset_change_event" in sql:
return _FakeResult([{"asset_id": 42, "change_type": "asset_modified"}])
if "FROM automation_operation_log" in sql:
return _FakeResult([])
raise AssertionError(f"unexpected SQL: {sql}")
class _FakeDbContext:
def __init__(self, db: _FakeLiveReadbackDb) -> None:
self.db = db
async def __aenter__(self) -> _FakeLiveReadbackDb:
return self.db
async def __aexit__(self, *_args: object) -> None:
return None
def test_live_change_event_readback_is_scoped_to_selected_asset_ids(
monkeypatch: Any,
) -> None:
now = datetime(2026, 7, 15, 1, 0, tzinfo=UTC)
db = _FakeLiveReadbackDb(now)
monkeypatch.setattr(
"src.db.base.get_db_context",
lambda _project_id: _FakeDbContext(db),
)
_latest_run, assets, _receipt = asyncio.run(_load_live_rows("awoooi"))
change_sql, change_params = next(
(sql, params) for sql, params in db.calls if "FROM asset_change_event" in sql
)
assert "asset_id = ANY(:asset_ids)" in change_sql
assert change_params == {"asset_ids": [42]}
assert assets[0]["change_types"] == ["asset_modified"]
def test_live_readback_timeout_fails_closed_without_hanging(
monkeypatch: Any,
) -> None:
async def _slow_live_readback(
_project_id: str,
) -> tuple[dict[str, Any], list[dict[str, Any]], dict[str, Any]]:
await asyncio.sleep(1)
return {}, [], {}
monkeypatch.setattr(matrix_service, "_load_live_rows", _slow_live_readback)
payload = asyncio.run(
build_asset_capability_matrix_with_live_readback(
repo_root=_REPO_ROOT,
timeout_seconds=0.01,
)
)
assert payload["live_readback_status"] == "degraded"
assert payload["live_readback_error_type"] == "TimeoutError"
assert payload["live_readback_timeout_seconds"] == 0.01
assert payload["summary"]["declared_asset_count"] == 193
assert payload["operation_boundaries"]["host_write_performed"] is False

View File

@@ -72,6 +72,35 @@
"container": "argocd-notifications-controller"
}
},
{
"asset_id": "runtime-image:argocd-repo-server",
"source_index_digest": "sha256:16b92ba472fbb9287459cc52e0ecff07288dff461209955098edb56ce866fe49",
"platform_digest": "sha256:2f25a42949ea69c0dd33f4ce1918c6a01039d6c14a7ecc1d19088504a9d3e94f",
"target_registry_digest": "sha256:8b7e25d7e6036d6750c5bda8920cbd80d8902c25771666f6767dc97bd7fab8fc",
"push_ref": "registry.wooo.work/awoooi/runtime-mirror/argocd:v3.3.6-linux-amd64",
"runtime_ref": "192.168.0.110:5000/awoooi/runtime-mirror/argocd@sha256:8b7e25d7e6036d6750c5bda8920cbd80d8902c25771666f6767dc97bd7fab8fc",
"workload": {
"kind": "deployment",
"namespace": "argocd",
"name": "argocd-repo-server",
"container": "argocd-repo-server"
}
},
{
"asset_id": "runtime-image:argocd-repo-server-copyutil",
"source_index_digest": "sha256:16b92ba472fbb9287459cc52e0ecff07288dff461209955098edb56ce866fe49",
"platform_digest": "sha256:2f25a42949ea69c0dd33f4ce1918c6a01039d6c14a7ecc1d19088504a9d3e94f",
"target_registry_digest": "sha256:8b7e25d7e6036d6750c5bda8920cbd80d8902c25771666f6767dc97bd7fab8fc",
"push_ref": "registry.wooo.work/awoooi/runtime-mirror/argocd:v3.3.6-linux-amd64",
"runtime_ref": "192.168.0.110:5000/awoooi/runtime-mirror/argocd@sha256:8b7e25d7e6036d6750c5bda8920cbd80d8902c25771666f6767dc97bd7fab8fc",
"workload": {
"kind": "deployment",
"namespace": "argocd",
"name": "argocd-repo-server",
"container": "copyutil",
"container_kind": "init_container"
}
},
{
"asset_id": "runtime-image:kured",
"source_index_digest": "sha256:2c5d73bb4517a269def38a6cd54d34d82be81793bea9ff1bb35c6533515ad209",

View File

@@ -160,12 +160,12 @@ spec:
- name: AWOOOI_BUILD_COMMIT_SHA
# 2026-06-29 Codex: CD rewrites this to the deployed image tag so
# production deploy readback does not rely on a stale static snapshot.
value: "b20d633aa25be07cf1faace558c39e6a39160c53"
value: "5a593fba5e110288185c5498509b43c37bc52859"
- name: AWOOOI_DESIRED_API_IMAGE_TAG
# 2026-06-30 Codex: CD rewrites this alongside AWOOOI_BUILD_COMMIT_SHA.
# Production readback compares runtime image truth against this
# GitOps desired tag instead of doing a slow Gitea raw fetch.
value: "b20d633aa25be07cf1faace558c39e6a39160c53"
value: "5a593fba5e110288185c5498509b43c37bc52859"
- name: DATABASE_POOL_SIZE
# 2026-07-01 Codex: production role `awoooi` currently has a low
# connection limit. Keep API pool conservative until DB role

View File

@@ -66,7 +66,7 @@ spec:
- name: DATABASE_NULL_POOL
value: "true"
- name: AWOOOI_BUILD_COMMIT_SHA
value: "b20d633aa25be07cf1faace558c39e6a39160c53"
value: "5a593fba5e110288185c5498509b43c37bc52859"
- name: ENABLE_AWOOOP_ANSIBLE_CANDIDATE_BACKFILL_WORKER
value: "false"
- name: ENABLE_AWOOOP_ANSIBLE_CHECK_MODE_WORKER

View File

@@ -173,7 +173,7 @@ spec:
- name: DATABASE_BOOTSTRAP_DDL_ENABLED
value: "false"
- name: AWOOOI_BUILD_COMMIT_SHA
value: "b20d633aa25be07cf1faace558c39e6a39160c53"
value: "5a593fba5e110288185c5498509b43c37bc52859"
- name: ENABLE_AWOOOP_ANSIBLE_CANDIDATE_BACKFILL_WORKER
value: "true"
- name: ENABLE_SECURITY_CONTROL_PLANE_MAINTENANCE_WORKER

View File

@@ -40,9 +40,9 @@ resources:
# ⚠️ 重要: name 必須與 deployment YAML 中的 image 完全匹配 (含 tag)
# newName + newTag 由 CI 透過 kustomize edit set image 注入
images:
- digest: sha256:7ef6d6dc72bf1737f7b69938d39705a39b94dc8a0f88067da6e0a0bab877676f
- digest: sha256:63787237ff5ace798b0978c0319969b9bff0a3cf44007dbd1c6eecadb4e9465e
name: 192.168.0.110:5000/library/api:IMAGE_TAG_PLACEHOLDER
newName: 192.168.0.110:5000/awoooi/api
- digest: sha256:3d02958c11ae2b2a2e05a9a8c0512465f98f46e8a8eb70596df0c90ffb7f7d3e
- digest: sha256:c454f0af821455d1b33947a706f3464ad06689cf6d3a681f0dcc5d420f90bad6
name: 192.168.0.110:5000/library/web:IMAGE_TAG_PLACEHOLDER
newName: 192.168.0.110:5000/awoooi/web

View File

@@ -47,7 +47,7 @@ class RuntimeImageMirrorControllerTest(unittest.TestCase):
self.assertEqual(policy.work_item_id, "AIA-P0-006-02A")
self.assertEqual(policy.risk_level, "high")
self.assertEqual(len(policy.images), 8)
self.assertEqual(len(policy.images), 10)
self.assertNotIn("github.com", raw)
self.assertNotIn("ghcr.io", raw)
self.assertIn('"external_pull_allowed": false', raw)
@@ -72,8 +72,8 @@ class RuntimeImageMirrorControllerTest(unittest.TestCase):
)
)
container_kinds = [image.workload.container_kind for image in policy.images]
self.assertEqual(container_kinds.count("container"), 7)
self.assertEqual(container_kinds.count("init_container"), 1)
self.assertEqual(container_kinds.count("container"), 8)
self.assertEqual(container_kinds.count("init_container"), 2)
def test_policy_loads_and_validates_init_container_kind(self) -> None:
image = self._image_with_container_kind("init_container")
@@ -184,6 +184,66 @@ class RuntimeImageMirrorControllerTest(unittest.TestCase):
image.runtime_ref.endswith("@" + image.target_registry_digest)
)
def test_argocd_repo_server_batch_reuses_verified_staging_artifact(self) -> None:
policy = controller.load_policy(POLICY_PATH)
staging = controller.load_staging_policy(STAGING_POLICY_PATH).images[0]
batch = {
image.asset_id: image
for image in policy.images
if image.asset_id
in {
"runtime-image:argocd-repo-server",
"runtime-image:argocd-repo-server-copyutil",
}
}
self.assertEqual(
set(batch),
{
"runtime-image:argocd-repo-server",
"runtime-image:argocd-repo-server-copyutil",
},
)
self.assertEqual(
{
(
image.workload.kind,
image.workload.namespace,
image.workload.name,
image.workload.container,
image.workload.container_kind,
)
for image in batch.values()
},
{
(
"deployment",
"argocd",
"argocd-repo-server",
"argocd-repo-server",
"container",
),
(
"deployment",
"argocd",
"argocd-repo-server",
"copyutil",
"init_container",
),
},
)
for image in batch.values():
self.assertEqual(image.source_index_digest, staging.source_index_digest)
self.assertEqual(image.platform_digest, staging.platform_digest)
self.assertEqual(image.push_ref, staging.push_ref)
self.assertEqual(
image.target_registry_digest,
"sha256:8b7e25d7e6036d6750c5bda8920cbd80d8902c25771666f6767dc97bd7fab8fc",
)
self.assertTrue(
image.runtime_ref.endswith("@" + image.target_registry_digest)
)
def test_policy_rejects_external_pull_and_mutable_runtime_ref(self) -> None:
payload = json.loads(POLICY_PATH.read_text(encoding="utf-8"))
with tempfile.TemporaryDirectory() as temp: