Files
aios/active-runtime/app/governed_production_release.py
T

589 lines
26 KiB
Python

"""Governed V2R2C1 production release orchestration.
This module is deliberately scoped to the Owner-approved V2R2C1 binding. It
uses the existing TagPublisher and metadata adapters; it does not introduce a
second Git writer, a second durable job store, or an alternate baseline.
"""
from __future__ import annotations
import argparse
import fcntl
import hashlib
import json
import os
import re
import subprocess
import tarfile
import tempfile
import time
from dataclasses import dataclass
from pathlib import Path
from typing import Any, Mapping
from gitea_tag_publication import (
GiteaTagClient,
GitTagBackend,
PublicationManualOwnerReviewRequired,
PublicationPending,
PublicationReconciliationRequired,
PublicationRequest,
TagPublisher,
)
VERSION = "V2R2C1"
TARGET_OID = "998d30304787e31e5ae607c0bd86b5edb2b0e5ad"
EVIDENCE_CLOSURE_OID = "a8b906d1bc24088c19647e98de6a679268f6d0a2"
APPROVAL_PLAN_ID = "sha256:c0df52008ae7f64cec5ee2d28947aa7b892066c670ab8295e4549140b1a7500e"
REPOSITORY = "aios/aios"
SERVICE = "aios-v2r1c7-production.service"
PORT = 18927
OID_RE = re.compile(r"^[0-9a-f]{40}$")
SHA_RE = re.compile(r"^[0-9a-f]{64}$")
class ReleaseBlocked(RuntimeError):
"""A release gate failed and no unsafe recovery is permitted."""
class ManualOwnerReviewRequired(ReleaseBlocked):
pass
class ReconciliationRequired(ReleaseBlocked):
pass
def canonical(value: object) -> bytes:
return (json.dumps(value, sort_keys=True, separators=(",", ":")) + "\n").encode()
def sha256_file(path: Path) -> str:
if path.is_symlink() or not path.is_file():
raise ReleaseBlocked("artifact must be a regular file")
digest = hashlib.sha256()
with path.open("rb") as stream:
for block in iter(lambda: stream.read(1024 * 1024), b""):
digest.update(block)
return digest.hexdigest()
def atomic_json(path: Path, value: Mapping[str, Any], mode: int = 0o600) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
temporary = path.with_name(path.name + ".tmp")
temporary.write_bytes(canonical(dict(value)))
os.chmod(temporary, mode)
os.replace(temporary, path)
@dataclass(frozen=True)
class ReleaseBinding:
version: str = VERSION
target_oid: str = TARGET_OID
evidence_closure_oid: str = EVIDENCE_CLOSURE_OID
approval_plan_id: str = APPROVAL_PLAN_ID
repository: str = REPOSITORY
production_service: str = SERVICE
production_port: int = PORT
def validate(self) -> None:
if self.version != VERSION or self.repository != REPOSITORY:
raise ManualOwnerReviewRequired("release binding is outside V2R2C1 scope")
if not OID_RE.fullmatch(self.target_oid) or self.target_oid != TARGET_OID:
raise ManualOwnerReviewRequired("V2R2C1 target_oid mismatch")
if not OID_RE.fullmatch(self.evidence_closure_oid) or self.evidence_closure_oid != EVIDENCE_CLOSURE_OID:
raise ManualOwnerReviewRequired("V2R2C1 evidence closure mismatch")
if self.approval_plan_id != APPROVAL_PLAN_ID:
raise ManualOwnerReviewRequired("V2R2C1 approval plan mismatch")
if self.production_service != SERVICE or int(self.production_port) != PORT:
raise ManualOwnerReviewRequired("production service binding mismatch")
def as_dict(self) -> dict[str, Any]:
return {
"version": self.version,
"target_oid": self.target_oid,
"evidence_closure_oid": self.evidence_closure_oid,
"approval_plan_id": self.approval_plan_id,
"repository": self.repository,
"production_service": self.production_service,
"production_port": self.production_port,
}
class StateLock:
def __init__(self, path: Path):
self.path = path
self.handle = None
def __enter__(self) -> "StateLock":
self.path.parent.mkdir(parents=True, exist_ok=True)
self.handle = self.path.open("a+", encoding="utf-8")
fcntl.flock(self.handle.fileno(), fcntl.LOCK_EX)
return self
def __exit__(self, exc_type, exc, tb) -> None:
if self.handle is not None:
fcntl.flock(self.handle.fileno(), fcntl.LOCK_UN)
self.handle.close()
class ReleaseState:
def __init__(self, root: Path, binding: ReleaseBinding):
self.root = root
self.path = root / "release-state.json"
self.binding = binding
root.mkdir(parents=True, exist_ok=True)
if self.path.exists():
value = json.loads(self.path.read_text(encoding="utf-8"))
if value.get("binding") != binding.as_dict():
raise ManualOwnerReviewRequired("persisted release binding conflict")
self.value = value
else:
self.value = {
"schema": "aios.v2r2c1.governed-production-release-state.v1",
"binding": binding.as_dict(),
"state": "INITIAL",
"completed": {},
"history": [],
}
self.save()
def save(self) -> None:
atomic_json(self.path, self.value)
def record(self, state: str, step: str, result: Mapping[str, Any]) -> None:
safe = json.loads(json.dumps(dict(result)))
self.value["state"] = state
self.value.setdefault("completed", {})[step] = safe
self.value.setdefault("history", []).append(
{"at": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()), "state": state, "step": step, "result": safe}
)
self.save()
def fail(self, state: str, reason: str) -> None:
self.value["state"] = state
self.value.setdefault("history", []).append(
{"at": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()), "state": state, "reason": str(reason)[:500]}
)
self.save()
def _git(git_dir: Path, *args: str, check: bool = True) -> str:
completed = subprocess.run(
["git", f"--git-dir={git_dir}", *args],
check=False,
capture_output=True,
text=True,
timeout=45,
env={"PATH": os.environ.get("PATH", "")},
)
if check and completed.returncode != 0:
raise ReconciliationRequired(completed.stderr.strip() or "authoritative git query failed")
return completed.stdout.strip()
def _ancestor(git_dir: Path, ancestor: str, descendant: str) -> bool:
return subprocess.run(
["git", f"--git-dir={git_dir}", "merge-base", "--is-ancestor", ancestor, descendant],
check=False,
capture_output=True,
timeout=45,
env={"PATH": os.environ.get("PATH", "")},
).returncode == 0
def verify_lineage(git_dir: Path, binding: ReleaseBinding) -> dict[str, Any]:
binding.validate()
if _git(git_dir, "cat-file", "-t", binding.target_oid) != "commit":
raise ReleaseBlocked("approved target_oid is not a commit")
if _git(git_dir, "cat-file", "-t", binding.evidence_closure_oid) != "commit":
raise ReleaseBlocked("evidence closure is not a commit")
if not _ancestor(git_dir, binding.target_oid, binding.evidence_closure_oid):
raise ManualOwnerReviewRequired("approved target is not an ancestor of evidence closure")
return {
"target_oid": binding.target_oid,
"evidence_closure_oid": binding.evidence_closure_oid,
"target_is_ancestor_of_evidence": True,
"main_oid": _git(git_dir, "rev-parse", "refs/heads/main"),
}
def verify_owner_approval(value: Mapping[str, Any], binding: ReleaseBinding) -> dict[str, Any]:
approval = value.get("owner_approval")
if not isinstance(approval, Mapping):
raise ReleaseBlocked("OWNER_APPROVAL_RECEIPT_REQUIRED")
if approval.get("approved") is not True or approval.get("status") != "APPROVED":
raise ReleaseBlocked("OWNER_APPROVAL_NOT_ACCEPTED")
if approval.get("approved_by") != "aiman":
raise ManualOwnerReviewRequired("OWNER_APPROVAL_ACTOR_MISMATCH")
if approval.get("approval_source") != "explicit_owner_approval_in_chat":
raise ManualOwnerReviewRequired("OWNER_APPROVAL_SOURCE_MISMATCH")
for key, expected in binding.as_dict().items():
if key in {"repository", "production_service", "production_port"}:
continue
if approval.get(key) != expected:
raise ManualOwnerReviewRequired("OWNER_APPROVAL_BINDING_MISMATCH:" + key)
return dict(approval)
def _payload_file(git_dir: Path, oid: str, path: str) -> bytes:
completed = subprocess.run(
["git", f"--git-dir={git_dir}", "show", f"{oid}:{path}"],
check=False,
capture_output=True,
timeout=45,
env={"PATH": os.environ.get("PATH", "")},
)
if completed.returncode != 0:
raise ReleaseBlocked("required release source missing:" + path)
return completed.stdout
def _commit_epoch(git_dir: Path, oid: str) -> int:
return int(_git(git_dir, "show", "-s", "--format=%ct", oid))
def build_production_artifact(
git_dir: Path,
binding: ReleaseBinding,
output_root: Path,
) -> dict[str, Any]:
"""Build a deterministic executable V2R2C1 health artifact from target_oid."""
binding.validate()
verify_lineage(git_dir, binding)
epoch = _commit_epoch(git_dir, binding.target_oid)
output_root.mkdir(parents=True, exist_ok=True)
stem = f"aios-{VERSION}-{binding.target_oid[:12]}"
output = output_root / (stem + ".tar.zst")
with tempfile.TemporaryDirectory(prefix="v2r2c1-artifact-") as temp:
root = Path(temp) / "payload"
root.mkdir(parents=True)
source_paths = {
"active-runtime/app/deployment_state.py": "active-runtime/app/deployment_state.py",
"active-runtime/app/v2r1c6_privacy_tools.py": "active-runtime/app/v2r1c6_privacy_tools.py",
}
for destination, source in source_paths.items():
target = root / destination
target.parent.mkdir(parents=True, exist_ok=True)
target.write_bytes(_payload_file(git_dir, binding.target_oid, source))
generated: dict[str, str] = {
"VERSION": VERSION + "\n",
"active-runtime/app/v2r2c1_health_service.py": (
"from v2r1c6_privacy_tools import main\n\n"
"if __name__ == '__main__':\n raise SystemExit(main())\n"
),
"deploy/aios-v2/systemd/aios-v2r1c7-production.service": (
"[Unit]\nDescription=AIOS V2R2C1 governed health contract service\n"
"After=network.target\n\n[Service]\nType=simple\n"
"User=aios-v2r1c7\nGroup=aios-v2r1c7\n"
"WorkingDirectory=/opt/aios/v2r2c1\n"
"EnvironmentFile=/etc/aios/aios-v2r2c1.env\n"
"ExecStart=/usr/bin/python3 /opt/aios/v2r2c1/active-runtime/app/v2r2c1_health_service.py --serve --host 127.0.0.1 --port ${AIOS_V2R1C7_PORT}\n"
"Restart=on-failure\nRestartSec=5\nNoNewPrivileges=true\nPrivateTmp=true\n"
"ProtectHome=true\nProtectSystem=strict\nProtectKernelLogs=true\n"
"ProtectKernelModules=true\nProtectKernelTunables=true\nProtectControlGroups=true\n"
"RestrictAddressFamilies=AF_UNIX AF_INET AF_INET6\nIPAddressDeny=any\n"
"IPAddressAllow=localhost\nUMask=0077\n\n[Install]\nWantedBy=multi-user.target\n"
),
"deploy/aios-v2/v2r2c1-runtime-config.json": json.dumps({
"schema": "aios.v2r2c1.production-runtime.v1",
"version": VERSION,
"target_oid": binding.target_oid,
"service": SERVICE,
"port": PORT,
"health_path": "/health",
"deployment_state_source": "environment_and_state_file",
}, sort_keys=True, separators=(",", ":")) + "\n",
"deploy/aios-v2/v2r2c1-rollback.json": json.dumps({
"schema": "aios.v2r2c1.rollback-metadata.v1",
"rollback_version": "V2R1C7",
"rollback_service": SERVICE,
"rollback_target_is_previous_production": True,
"automatic_rollback": False,
}, sort_keys=True, separators=(",", ":")) + "\n",
"scripts/probe_v2r2c1_health.py": (
"import json,sys,urllib.request\n"
"url=sys.argv[1] if len(sys.argv)>1 else 'http://127.0.0.1:18927/health'\n"
"with urllib.request.urlopen(url, timeout=5) as r: print(json.dumps(json.load(r), sort_keys=True))\n"
),
"scripts/rollback_v2r2c1_release.sh": "#!/bin/sh\nset -eu\nprintf '%s\\n' 'rollback requires governed owner-approved executor'\nexit 2\n",
"release-manifests/V2R2C1.json": json.dumps({
"schema": "aios.release-source.v2",
"version": VERSION,
"repository": REPOSITORY,
"target_oid": binding.target_oid,
"evidence_closure_oid": binding.evidence_closure_oid,
"production_deployed": False,
"deployment_state": "STAGING",
"artifact_kind": "executable-production-health",
}, sort_keys=True, separators=(",", ":")) + "\n",
}
for rel, text in generated.items():
target = root / rel
target.parent.mkdir(parents=True, exist_ok=True)
target.write_text(text, encoding="utf-8")
if rel.startswith("scripts/"):
target.chmod(0o755)
metadata = {
"schema": "aios.release-artifact-build.v3",
"version": VERSION,
"repository": REPOSITORY,
"source_head_oid": binding.target_oid,
"evidence_closure_oid": binding.evidence_closure_oid,
"source_date_epoch": epoch,
"clean_source": True,
"executable_payload": True,
"production_deployed": False,
"tag_created": False,
"baseline_changed": False,
"secrets_included": False,
}
(root / "BUILD-METADATA.json").write_bytes(canonical(metadata))
(root / "ROLLBACK-METADATA.json").write_bytes(canonical(json.loads((root / "deploy/aios-v2/v2r2c1-rollback.json").read_text(encoding="utf-8"))))
names = sorted("payload/" + str(path.relative_to(root)).replace(os.sep, "/") for path in root.rglob("*") if path.is_file())
(root / "ARTIFACT-MANIFEST.json").write_bytes(canonical({
"schema": "aios.release-artifact-manifest.v3",
"version": VERSION,
"source_head_oid": binding.target_oid,
"evidence_closure_oid": binding.evidence_closure_oid,
"repository": REPOSITORY,
"files": names + ["payload/ARTIFACT-MANIFEST.json"],
"deployable": True,
"executable_payload": True,
"production_deployed": False,
"tag_created": False,
"baseline_changed": False,
}))
tar_path = Path(temp) / "release.tar"
with tarfile.open(tar_path, "w") as archive:
for path in sorted(root.rglob("*")):
relative = str(path.relative_to(root)).replace(os.sep, "/")
info = archive.gettarinfo(str(path), arcname="payload/" + relative)
info.uid = info.gid = 0
info.uname = "root"
info.gname = "root"
info.mtime = epoch
info.mode = 0o755 if path.is_dir() or path.suffix == ".sh" else 0o644
if path.is_file():
with path.open("rb") as stream:
archive.addfile(info, stream)
else:
archive.addfile(info)
subprocess.run(["zstd", "-T1", "-19", "--no-progress", "-q", "-f", str(tar_path), "-o", str(output)], check=True, timeout=120)
digest = sha256_file(output)
receipt = {
"schema": "aios.release-artifact-receipt.v3",
"version": VERSION,
"repository": REPOSITORY,
"source_head_oid": binding.target_oid,
"evidence_closure_oid": binding.evidence_closure_oid,
"artifact": str(output),
"artifact_sha256": digest,
"artifact_size": output.stat().st_size,
"deployable": True,
"executable_payload": True,
"production_deployed": False,
"tag_created": False,
"baseline_changed": False,
}
atomic_json(output.with_suffix(output.suffix + ".receipt.json"), receipt, mode=0o640)
return receipt
class GovernedProductionReleaseExecutor:
def __init__(self, config: Any):
self.config = config
self.binding = ReleaseBinding()
self.binding.validate()
self.root = Path(config.state_root) / "v2r2c1-governed-production-release"
self.state = ReleaseState(self.root, self.binding)
workspace = self.root / "publication-workspace"
workspace.mkdir(parents=True, exist_ok=True)
self.publisher = TagPublisher(
workspace,
GitTagBackend(workspace, authoritative_repo=Path(config.authoritative_git_dir)),
GiteaTagClient(config.api_base, REPOSITORY, Path(config.token_path)),
config.remote_name,
state_root=self.root / "tag-publication",
)
# Release checks must use the same explicitly configured staging baseline
# as the Action/model runtime; do not fall back to the publication workspace.
self.publisher.baseline_path = Path(
os.environ.get("AIOS_GITOPS_BASELINE_AUTHORITY", str(config.baseline_path))
)
# The immutable runtime package is the source of the version projection;
# the mutable state workspace only stores publication state.
self.publisher.versions_root = Path(__file__).resolve().parents[1] / "versions"
def _request(self, request: Mapping[str, Any]) -> PublicationRequest:
binding = self.binding
for key, expected in binding.as_dict().items():
if key in request and request[key] != expected:
raise ManualOwnerReviewRequired("release request binding mismatch:" + key)
return PublicationRequest(binding.version, binding.target_oid, binding.evidence_closure_oid, binding.approval_plan_id)
def approval_request(self, request: Mapping[str, Any]) -> dict[str, Any]:
if request.get("execute") is not False:
raise ReleaseBlocked("TAG_APPROVAL_REQUEST_MUST_BE_DRY_RUN")
owner = verify_owner_approval(request, self.binding)
with StateLock(self.root / "release.lock"):
current = self.state.value.get("approval")
approval_digest = hashlib.sha256(canonical(owner)).hexdigest()
record = {
"schema": "aios.v2r2c1.tag-approval-request.v1",
"binding": self.binding.as_dict(),
"owner_approval": owner,
"owner_approval_digest_sha256": approval_digest,
"execute": False,
"recorded_at": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()),
}
if current and current != record:
raise ManualOwnerReviewRequired("persisted Owner approval differs")
self.state.value["approval"] = record
self.state.record("APPROVAL_REQUEST_RECORDED", "tag_approval_request", record)
return {"status": "RECORDED", "execute": False, "binding": self.binding.as_dict(), "owner_approval_digest_sha256": approval_digest}
def _require_approval(self) -> None:
approval = self.state.value.get("approval")
if not isinstance(approval, Mapping) or approval.get("binding") != self.binding.as_dict():
raise ReleaseBlocked("OWNER_APPROVAL_REQUEST_NOT_RECORDED")
def publish_tag(self, request: Mapping[str, Any]) -> dict[str, Any]:
if request.get("execute") is not True:
raise ReleaseBlocked("TAG_PUBLICATION_REQUIRES_EXECUTE_TRUE")
self._require_approval()
publication = self._request(request)
with StateLock(self.root / "release.lock"):
try:
verified = self.publisher._verify_publication(publication, allow_push=True, previously_deployed=False)
except PublicationManualOwnerReviewRequired as exc:
self.state.fail("MANUAL_OWNER_REVIEW_REQUIRED", str(exc))
raise ManualOwnerReviewRequired(str(exc)) from exc
except PublicationPending as exc:
self.state.fail("REMOTE_PUSH_PENDING", str(exc))
raise ReconciliationRequired(str(exc)) from exc
except PublicationReconciliationRequired as exc:
self.state.fail("RECONCILIATION_REQUIRED", str(exc))
raise ReconciliationRequired(str(exc)) from exc
result = dict(verified)
self.state.record("TAG_VERIFIED", "tag", result)
return {"status": "REMOTE_VERIFIED", **result}
def tag_status(self, request: Mapping[str, Any]) -> dict[str, Any]:
self._request(request)
return self.publisher.get_tag_publication_status(VERSION)
def build_artifact(self, request: Mapping[str, Any]) -> dict[str, Any]:
self._require_approval()
self._request(request)
receipt = build_production_artifact(Path(self.config.authoritative_git_dir), self.binding, self.root / "artifacts")
self.state.record("ARTIFACT_VERIFIED", "production_artifact", receipt)
return {"status": "VERIFIED", **receipt}
def execute_request(self, request: Mapping[str, Any]) -> dict[str, Any]:
self._require_approval()
publication = self._request(request)
tag_state = self.state.value.get("completed", {}).get("tag")
if not isinstance(tag_state, Mapping) or tag_state.get("state") not in {"GITEA_API_CONFIRMED", "REMOTE_VERIFIED"}:
raise ReleaseBlocked("REMOTE_TAG_PROOF_REQUIRED")
artifact = self.state.value.get("completed", {}).get("production_artifact")
if not isinstance(artifact, Mapping):
artifact = self.build_artifact(request)
value = {
**self.binding.as_dict(),
"tag_oid": tag_state.get("tag_oid") or tag_state.get("gitea_tag_oid"),
"artifact_path": artifact["artifact"],
"artifact_size": int(artifact["artifact_size"]),
"artifact_sha256": artifact["artifact_sha256"],
}
# The production adapter is privileged and is run through the root
# approved-release service boundary by the caller. This handler is
# still useful for a root-owned staging invocation and remains
# fail-closed when the adapter is not available.
import v2r2c1_production_adapter
runtime = v2r2c1_production_adapter.execute_request({"schema": "aios.production-executor-request.v1", "operation": "apply", **value})
if runtime.get("status") != "ACTIVE_VERIFIED":
self.state.fail("RECONCILIATION_REQUIRED", "production health did not verify")
raise ReconciliationRequired("production health did not verify")
self.state.record("PRODUCTION_VERIFIED", "production", runtime)
return {"status": "PRODUCTION_VERIFIED", "runtime": runtime, "artifact": artifact, "tag_oid": value["tag_oid"]}
def consistency(self, request: Mapping[str, Any]) -> dict[str, Any]:
publication = self._request(request)
result = self.publisher.get_release_consistency(publication)
result["release_binding"] = self.binding.as_dict()
result["production_health"] = _health(self.config.health_url, self.binding, self.state.value.get("completed", {}).get("production_artifact", {}))
result["registry"] = _read_json(Path(self.config.registry_root) / f"{VERSION}.json")
result["baseline"] = _read_json(Path(self.config.baseline_path))
return result
def acceptance(self, request: Mapping[str, Any]) -> dict[str, Any]:
result = self.consistency(request)
result["accepted"] = (
result.get("status") == "PASS"
and result.get("production_health", {}).get("verification_status")
== "ACTIVE_VERIFIED"
)
self.state.record("ACCEPTED" if result["accepted"] else "RECONCILIATION_REQUIRED", "acceptance", result)
return result
def _read_json(path: Path) -> dict[str, Any]:
try:
value = json.loads(path.read_text(encoding="utf-8"))
except FileNotFoundError:
return {"status": "MISSING", "path": str(path)}
if not isinstance(value, dict):
return {"status": "INVALID", "path": str(path)}
return {"status": "PRESENT", "path": str(path), **value}
def _health(url: str, binding: ReleaseBinding, artifact: Mapping[str, Any]) -> dict[str, Any]:
import urllib.request
try:
with urllib.request.urlopen(url, timeout=5) as response:
body = json.loads(response.read().decode("utf-8"))
status = response.status
except Exception as exc:
return {
"status": "UNAVAILABLE",
"verification_status": "UNAVAILABLE",
"error": type(exc).__name__,
}
expected_artifact = artifact.get("artifact_sha256")
exact = (
status == 200
and body.get("version") == binding.version
and body.get("commit_oid") == binding.target_oid
and (not expected_artifact or body.get("artifact_sha256") == expected_artifact)
and body.get("deployment_state") == "ACTIVE"
and body.get("production_deployed") is True
)
return {
**body,
"http_status": status,
"verification_status": "ACTIVE_VERIFIED" if exact else "CONFLICT",
}
def build_executor_config(config: Any) -> GovernedProductionReleaseExecutor:
return GovernedProductionReleaseExecutor(config)
def main() -> int:
parser = argparse.ArgumentParser()
parser.add_argument("operation", choices=("build-artifact", "status"))
parser.add_argument("--git-dir", type=Path, default=Path("/data/aios-gitops/repo.git"))
parser.add_argument("--output-root", type=Path, default=Path("/var/lib/aios-gitops/approved-artifacts/V2R2C1"))
args = parser.parse_args()
binding = ReleaseBinding()
if args.operation == "build-artifact":
print(json.dumps(build_production_artifact(args.git_dir, binding, args.output_root), sort_keys=True))
else:
print(json.dumps({"binding": binding.as_dict(), "status": "READY"}, sort_keys=True))
return 0
if __name__ == "__main__":
raise SystemExit(main())