feat: add secure agent enrollment

This commit is contained in:
lussierj committed 2026-08-26 21:31:09 +00:00
1 parent 82309db79a
commit d9f0da694b
26 files changed
+2942 -76

No files matched your search

+1
View File
@@ -0,0 +1 @@
"""Secure one-time Agent enrollment."""
@@ -0,0 +1,8 @@
"""Dependency wiring for secure Agent enrollment."""
from app.agent_enrollment.repository import AgentEnrollmentRepository
from app.agent_enrollment.service import AgentEnrollmentService
def get_agent_enrollment_service() -> AgentEnrollmentService:
return AgentEnrollmentService(AgentEnrollmentRepository())
@@ -0,0 +1,21 @@
"""Domain exceptions for secure Agent enrollment."""
class AgentEnrollmentError(Exception):
"""Base Agent enrollment error."""
class AgentEnrollmentNotFound(AgentEnrollmentError):
"""The requested Site or Agent Zone does not exist."""
class AgentEnrollmentConflict(AgentEnrollmentError):
"""The requested enrollment conflicts with current Agent routing state."""
class AgentEnrollmentTokenInvalid(AgentEnrollmentError):
"""The supplied bootstrap token cannot be redeemed."""
class AgentEnrollmentCredentialCollision(AgentEnrollmentError):
"""A generated machine credential collided with an existing credential."""
+94
View File
@@ -0,0 +1,94 @@
"""Typed contracts for secure Agent enrollment."""
from datetime import datetime
from typing import Literal
from uuid import UUID
from pydantic import BaseModel, ConfigDict, Field, field_validator
class AgentEnrollmentCreateRequest(BaseModel):
model_config = ConfigDict(extra="forbid")
site_id: UUID
agent_zone_id: UUID
display_name: str = Field(min_length=1, max_length=120)
@field_validator("display_name")
@classmethod
def normalize_display_name(cls, value: str) -> str:
normalized = value.strip()
if not normalized:
raise ValueError("display_name cannot be blank")
return normalized
class AgentEnrollmentCreated(BaseModel):
id: UUID
site_id: UUID
site_name: str
agent_zone_id: UUID
agent_zone_name: str
display_name: str
created_at: datetime
expires_at: datetime
enrollment_token: str
class AgentEnrollmentCreateResponse(BaseModel):
data: AgentEnrollmentCreated
meta: dict[str, object] = Field(default_factory=dict)
error: None = None
class AgentEnrollmentSummary(BaseModel):
id: UUID
site_id: UUID
site_name: str
agent_zone_id: UUID
agent_zone_name: str
display_name: str
created_by_user_id: UUID | None = None
created_at: datetime
expires_at: datetime
used_at: datetime | None = None
revoked_at: datetime | None = None
enrolled_agent_id: str | None = None
status: Literal["waiting", "used", "revoked", "expired"]
class AgentEnrollmentListResponse(BaseModel):
data: list[AgentEnrollmentSummary]
meta: dict[str, object] = Field(default_factory=dict)
error: None = None
class AgentEnrollRequest(BaseModel):
model_config = ConfigDict(extra="forbid")
token: str = Field(min_length=32, max_length=256)
@field_validator("token")
@classmethod
def normalize_token(cls, value: str) -> str:
normalized = value.strip()
if not normalized:
raise ValueError("token cannot be blank")
return normalized
class AgentEnrollCredentials(BaseModel):
agent_id: str
agent_secret: str
site_id: UUID
site_name: str
agent_zone_id: UUID
agent_zone_name: str
display_name: str
ip_ranges: list[str] = Field(default_factory=list)
class AgentEnrollResponse(BaseModel):
data: AgentEnrollCredentials
meta: dict[str, object] = Field(default_factory=dict)
error: None = None
+506
View File
@@ -0,0 +1,506 @@
"""PostgreSQL persistence for secure Agent enrollment."""
from collections.abc import Callable
from uuid import UUID
import psycopg
from psycopg.errors import UniqueViolation
from psycopg.rows import dict_row
from psycopg.types.json import Jsonb
from app.agent_enrollment.exceptions import (
AgentEnrollmentConflict,
AgentEnrollmentCredentialCollision,
AgentEnrollmentNotFound,
AgentEnrollmentTokenInvalid,
)
from app.config import settings
ConnectionFactory = Callable[[], psycopg.Connection]
TOKEN_HASH_CONSTRAINT = "uq_agent_enrollments_token_hash"
OPEN_ZONE_CONSTRAINT = "uq_agent_enrollments_open_zone"
AGENT_SECRET_HASH_CONSTRAINT = "uq_minerops_site_agents_agent_secret_hash"
AGENT_ZONE_CONSTRAINT = "uq_minerops_site_agents_agent_zone_id"
AGENT_ID_CONSTRAINT = "minerops_site_agents_pkey"
def default_connection_factory() -> psycopg.Connection:
return psycopg.connect(
host=settings.postgres_host,
port=settings.postgres_port,
dbname=settings.postgres_db,
user=settings.postgres_user,
password=settings.postgres_password,
connect_timeout=5,
)
class AgentEnrollmentRepository:
def __init__(self, connection_factory: ConnectionFactory | None = None):
self._connection_factory = connection_factory or default_connection_factory
def create_enrollment(
self,
*,
token_hash: str,
site_id: UUID,
agent_zone_id: UUID,
display_name: str,
actor_user_id: UUID,
) -> dict:
connection = self._connection_factory()
try:
with connection.cursor(row_factory=dict_row) as cur:
cur.execute(
"""
SELECT zone.id,
zone.site_id,
zone.name AS agent_zone_name,
zone.is_enabled,
site.name AS site_name,
agent.agent_id AS assigned_agent_id
FROM agent_zones AS zone
JOIN sites AS site ON site.id = zone.site_id
LEFT JOIN minerops_site_agents AS agent
ON agent.agent_zone_id = zone.id
WHERE zone.id = %s
AND zone.site_id = %s
FOR UPDATE OF zone
""",
(agent_zone_id, site_id),
)
zone = cur.fetchone()
if not zone:
raise AgentEnrollmentNotFound(
"Agent Zone was not found for the selected Site"
)
if zone["is_enabled"] is not True:
raise AgentEnrollmentConflict(
"Agent Zone is disabled"
)
if zone["assigned_agent_id"]:
raise AgentEnrollmentConflict(
"Agent Zone already has an assigned Agent"
)
cur.execute(
"SELECT clock_timestamp() AS operation_now"
)
operation_now = cur.fetchone()["operation_now"]
cur.execute(
"""
UPDATE agent_enrollments
SET revoked_at = %s
WHERE agent_zone_id = %s
AND used_at IS NULL
AND revoked_at IS NULL
RETURNING id
""",
(operation_now, agent_zone_id),
)
replaced_ids = [
str(row["id"]) for row in cur.fetchall()
]
cur.execute(
"""
INSERT INTO agent_enrollments (
token_hash,
site_id,
agent_zone_id,
display_name,
created_by_user_id,
created_at,
expires_at
)
VALUES (
%s,
%s,
%s,
%s,
%s,
%s,
%s + interval '30 minutes'
)
RETURNING
id,
site_id,
agent_zone_id,
display_name,
created_at,
expires_at
""",
(
token_hash,
site_id,
agent_zone_id,
display_name,
actor_user_id,
operation_now,
operation_now,
),
)
enrollment = dict(cur.fetchone())
enrollment["site_name"] = zone["site_name"]
enrollment["agent_zone_name"] = zone["agent_zone_name"]
cur.execute(
"""
INSERT INTO audit_logs (
actor_user_id,
action,
entity_type,
entity_id,
new_value_json,
created_at
)
VALUES (
%s,
'agent.enrollment.created',
'agent_enrollment',
%s,
%s,
%s
)
""",
(
actor_user_id,
enrollment["id"],
Jsonb(
{
"site_id": str(site_id),
"agent_zone_id": str(agent_zone_id),
"display_name": display_name,
"expires_at": enrollment[
"expires_at"
].isoformat(),
"replaced_enrollment_ids": replaced_ids,
}
),
operation_now,
),
)
connection.commit()
return enrollment
except UniqueViolation as exc:
connection.rollback()
constraint = exc.diag.constraint_name
if constraint == TOKEN_HASH_CONSTRAINT:
raise AgentEnrollmentCredentialCollision(
"Generated enrollment token collided"
) from exc
if constraint == OPEN_ZONE_CONSTRAINT:
raise AgentEnrollmentConflict(
"An active enrollment already exists for this Agent Zone"
) from exc
raise
except Exception:
connection.rollback()
raise
finally:
connection.close()
def list_enrollments(
self,
*,
site_id: UUID | None = None,
agent_zone_id: UUID | None = None,
limit: int = 100,
) -> list[dict]:
connection = self._connection_factory()
try:
with connection.cursor(row_factory=dict_row) as cur:
cur.execute(
"""
SELECT
enrollment.id,
enrollment.site_id,
site.name AS site_name,
enrollment.agent_zone_id,
zone.name AS agent_zone_name,
enrollment.display_name,
enrollment.created_by_user_id,
enrollment.created_at,
enrollment.expires_at,
enrollment.used_at,
enrollment.revoked_at,
enrollment.enrolled_agent_id,
CASE
WHEN enrollment.used_at IS NOT NULL
THEN 'used'
WHEN enrollment.revoked_at IS NOT NULL
THEN 'revoked'
WHEN enrollment.expires_at <= now()
THEN 'expired'
ELSE 'waiting'
END AS status
FROM agent_enrollments AS enrollment
JOIN sites AS site
ON site.id = enrollment.site_id
JOIN agent_zones AS zone
ON zone.id = enrollment.agent_zone_id
WHERE (%s::uuid IS NULL OR enrollment.site_id = %s::uuid)
AND (
%s::uuid IS NULL
OR enrollment.agent_zone_id = %s::uuid
)
ORDER BY enrollment.created_at DESC, enrollment.id DESC
LIMIT %s
""",
(
site_id,
site_id,
agent_zone_id,
agent_zone_id,
limit,
),
)
return [dict(row) for row in cur.fetchall()]
finally:
connection.close()
def redeem_enrollment(
self,
*,
token_hash: str,
agent_id: str,
agent_secret_hash: str,
) -> dict:
connection = self._connection_factory()
try:
with connection.cursor(row_factory=dict_row) as cur:
cur.execute(
"""
SELECT
id,
site_id,
agent_zone_id,
display_name,
expires_at,
used_at,
revoked_at
FROM agent_enrollments
WHERE token_hash = %s
FOR UPDATE
""",
(token_hash,),
)
enrollment = cur.fetchone()
if not enrollment:
raise AgentEnrollmentTokenInvalid(
"Invalid or expired enrollment token"
)
if (
enrollment["used_at"] is not None
or enrollment["revoked_at"] is not None
):
raise AgentEnrollmentTokenInvalid(
"Invalid or expired enrollment token"
)
cur.execute(
"""
SELECT
zone.id,
zone.site_id,
zone.name AS agent_zone_name,
zone.is_enabled,
site.name AS site_name
FROM agent_zones AS zone
JOIN sites AS site ON site.id = zone.site_id
WHERE zone.id = %s
AND zone.site_id = %s
FOR UPDATE OF zone
""",
(
enrollment["agent_zone_id"],
enrollment["site_id"],
),
)
zone = cur.fetchone()
if not zone:
raise AgentEnrollmentConflict(
"Enrollment Agent Zone no longer exists"
)
if zone["is_enabled"] is not True:
raise AgentEnrollmentConflict(
"Agent Zone is disabled"
)
cur.execute(
"""
SELECT agent_id
FROM minerops_site_agents
WHERE agent_zone_id = %s
FOR UPDATE
""",
(enrollment["agent_zone_id"],),
)
existing_agent = cur.fetchone()
if existing_agent:
raise AgentEnrollmentConflict(
"Agent Zone already has an assigned Agent"
)
cur.execute(
"SELECT clock_timestamp() AS operation_now"
)
operation_now = cur.fetchone()["operation_now"]
if enrollment["expires_at"] <= operation_now:
raise AgentEnrollmentTokenInvalid(
"Invalid or expired enrollment token"
)
cur.execute(
"""
INSERT INTO minerops_site_agents (
agent_id,
site_id,
agent_zone_id,
site_name,
display_name,
agent_secret_hash,
credential_created_at
)
VALUES (
%s,
%s,
%s,
%s,
%s,
%s,
%s
)
""",
(
agent_id,
enrollment["site_id"],
enrollment["agent_zone_id"],
zone["site_name"],
enrollment["display_name"],
agent_secret_hash,
operation_now,
),
)
cur.execute(
"""
UPDATE agent_enrollments
SET used_at = %s,
enrolled_agent_id = %s
WHERE id = %s
RETURNING used_at
""",
(operation_now, agent_id, enrollment["id"]),
)
used_at = cur.fetchone()["used_at"]
cur.execute(
"""
SELECT network::text AS network
FROM agent_zone_network_ranges
WHERE agent_zone_id = %s
ORDER BY network, id
""",
(enrollment["agent_zone_id"],),
)
networks = [
row["network"] for row in cur.fetchall()
]
cur.execute(
"""
INSERT INTO audit_logs (
actor_user_id,
action,
entity_type,
entity_id,
new_value_json,
created_at
)
VALUES (
NULL,
'agent.enrollment.redeemed',
'agent_enrollment',
%s,
%s,
%s
)
""",
(
enrollment["id"],
Jsonb(
{
"agent_id": agent_id,
"site_id": str(enrollment["site_id"]),
"agent_zone_id": str(
enrollment["agent_zone_id"]
),
"display_name": enrollment[
"display_name"
],
"used_at": used_at.isoformat(),
}
),
operation_now,
),
)
result = {
"agent_id": agent_id,
"site_id": enrollment["site_id"],
"site_name": zone["site_name"],
"agent_zone_id": enrollment["agent_zone_id"],
"agent_zone_name": zone["agent_zone_name"],
"display_name": enrollment["display_name"],
"ip_ranges": networks,
}
connection.commit()
return result
except UniqueViolation as exc:
connection.rollback()
constraint = exc.diag.constraint_name
if constraint in {
AGENT_ID_CONSTRAINT,
AGENT_SECRET_HASH_CONSTRAINT,
}:
raise AgentEnrollmentCredentialCollision(
"Generated Agent credential collided"
) from exc
if constraint == AGENT_ZONE_CONSTRAINT:
raise AgentEnrollmentConflict(
"Agent Zone already has an assigned Agent"
) from exc
raise
except Exception:
connection.rollback()
raise
finally:
connection.close()
+108
View File
@@ -0,0 +1,108 @@
"""HTTP routes for Agent enrollment and bootstrap."""
from typing import Annotated
from uuid import UUID
from fastapi import APIRouter, Depends, HTTPException, Query, Response, status
from app.agent_enrollment.dependencies import get_agent_enrollment_service
from app.agent_enrollment.exceptions import (
AgentEnrollmentConflict,
AgentEnrollmentNotFound,
AgentEnrollmentTokenInvalid,
)
from app.agent_enrollment.models import (
AgentEnrollmentCreateRequest,
AgentEnrollmentCreateResponse,
AgentEnrollmentListResponse,
AgentEnrollRequest,
AgentEnrollResponse,
)
from app.agent_enrollment.service import AgentEnrollmentService
from app.identity import require_permission
router = APIRouter(tags=["agent-enrollment"])
EnrollmentService = Annotated[
AgentEnrollmentService,
Depends(get_agent_enrollment_service),
]
EnrollmentViewer = Annotated[
dict,
Depends(require_permission("agents.view")),
]
EnrollmentManager = Annotated[
dict,
Depends(require_permission("agents.manage")),
]
@router.post(
"/api/agent-enrollments",
response_model=AgentEnrollmentCreateResponse,
status_code=status.HTTP_201_CREATED,
)
def create_agent_enrollment(
request: AgentEnrollmentCreateRequest,
response: Response,
current_user: EnrollmentManager,
service: EnrollmentService,
) -> AgentEnrollmentCreateResponse:
response.headers["Cache-Control"] = "no-store"
try:
enrollment = service.create_enrollment(
request,
actor_user_id=UUID(str(current_user["id"])),
)
return AgentEnrollmentCreateResponse(data=enrollment)
except AgentEnrollmentNotFound as exc:
raise HTTPException(status_code=404, detail=str(exc)) from exc
except AgentEnrollmentConflict as exc:
raise HTTPException(status_code=409, detail=str(exc)) from exc
@router.get(
"/api/agent-enrollments",
response_model=AgentEnrollmentListResponse,
)
def list_agent_enrollments(
current_user: EnrollmentViewer,
service: EnrollmentService,
site_id: UUID | None = Query(default=None),
agent_zone_id: UUID | None = Query(default=None),
limit: int = Query(default=100, ge=1, le=500),
) -> AgentEnrollmentListResponse:
del current_user
return AgentEnrollmentListResponse(
data=service.list_enrollments(
site_id=site_id,
agent_zone_id=agent_zone_id,
limit=limit,
)
)
@router.post(
"/api/agent/enroll",
response_model=AgentEnrollResponse,
status_code=status.HTTP_201_CREATED,
)
def enroll_agent(
request: AgentEnrollRequest,
response: Response,
service: EnrollmentService,
) -> AgentEnrollResponse:
response.headers["Cache-Control"] = "no-store"
try:
return AgentEnrollResponse(data=service.redeem(request))
except AgentEnrollmentTokenInvalid as exc:
raise HTTPException(
status_code=401,
detail="Invalid or expired enrollment token",
) from exc
except AgentEnrollmentConflict as exc:
raise HTTPException(status_code=409, detail=str(exc)) from exc
+98
View File
@@ -0,0 +1,98 @@
"""Application service for secure one-time Agent enrollment."""
import secrets
from uuid import UUID, uuid4
from app.agent_auth.service import hash_agent_secret
from app.agent_enrollment.exceptions import (
AgentEnrollmentConflict,
AgentEnrollmentCredentialCollision,
)
from app.agent_enrollment.models import (
AgentEnrollmentCreateRequest,
AgentEnrollRequest,
)
MAX_GENERATION_ATTEMPTS = 3
def generate_enrollment_token() -> str:
return secrets.token_urlsafe(32)
def generate_agent_secret() -> str:
return secrets.token_urlsafe(32)
def generate_agent_id() -> str:
return f"agent-{uuid4().hex}"
class AgentEnrollmentService:
def __init__(self, repository):
self.repository = repository
def create_enrollment(
self,
request: AgentEnrollmentCreateRequest,
*,
actor_user_id: UUID,
) -> dict:
for _ in range(MAX_GENERATION_ATTEMPTS):
token = generate_enrollment_token()
try:
enrollment = self.repository.create_enrollment(
token_hash=hash_agent_secret(token),
site_id=request.site_id,
agent_zone_id=request.agent_zone_id,
display_name=request.display_name,
actor_user_id=actor_user_id,
)
return {
**enrollment,
"enrollment_token": token,
}
except AgentEnrollmentCredentialCollision:
continue
raise AgentEnrollmentConflict(
"Unable to generate a unique enrollment token"
)
def list_enrollments(
self,
*,
site_id: UUID | None = None,
agent_zone_id: UUID | None = None,
limit: int = 100,
) -> list[dict]:
return self.repository.list_enrollments(
site_id=site_id,
agent_zone_id=agent_zone_id,
limit=limit,
)
def redeem(self, request: AgentEnrollRequest) -> dict:
token_hash = hash_agent_secret(request.token)
for _ in range(MAX_GENERATION_ATTEMPTS):
agent_id = generate_agent_id()
agent_secret = generate_agent_secret()
try:
result = self.repository.redeem_enrollment(
token_hash=token_hash,
agent_id=agent_id,
agent_secret_hash=hash_agent_secret(agent_secret),
)
return {
**result,
"agent_secret": agent_secret,
}
except AgentEnrollmentCredentialCollision:
continue
raise AgentEnrollmentConflict(
"Unable to generate unique Agent credentials"
)