feat(telemetry): add portable local telemetry plane
This commit is contained in:
DCCONSTRUCTIONS
@@ -0,0 +1,10 @@
|
||||
MISSIONCORE_MQTT_BIND_ADDRESS=127.0.0.1
|
||||
MISSIONCORE_MQTT_PORT=1883
|
||||
MISSIONCORE_TELEMETRY_QUERY_PORT=18030
|
||||
MISSIONCORE_DB_NAME=missioncore_telemetry
|
||||
MISSIONCORE_DB_USER=missioncore_ingest
|
||||
MISSIONCORE_DB_PASSWORD=replace-with-a-random-local-secret
|
||||
MISSIONCORE_MQTT_INGEST_USER=missioncore-ingest
|
||||
MISSIONCORE_MQTT_INGEST_PASSWORD=replace-with-a-random-local-secret
|
||||
MISSIONCORE_MQTT_WORKER_006_USER=worker-006
|
||||
MISSIONCORE_MQTT_WORKER_006_PASSWORD=replace-with-a-different-random-local-secret
|
||||
@@ -0,0 +1,2 @@
|
||||
.env
|
||||
runtime/
|
||||
@@ -0,0 +1,81 @@
|
||||
# NDC Mission Core local telemetry plane
|
||||
|
||||
Portable local-only telemetry infrastructure for compute contours.
|
||||
|
||||
One Docker Compose project, `ndc-mission-core-telemetry`, owns three isolated
|
||||
containers:
|
||||
|
||||
- `ndc-mission-core-mqtt-broker` — Eclipse Mosquitto `2.1.2-alpine`;
|
||||
- `ndc-mission-core-telemetry-timescaledb` — TimescaleDB HA OSS
|
||||
`pg16.14-ts2.28.2-all-oss`;
|
||||
- `ndc-mission-core-telemetry-normalizer` — Mission Core telemetry normalizer.
|
||||
|
||||
The Telegraf configuration templates cover Windows and Linux, but the agent runs as a
|
||||
host service rather than inside this Compose project. This preserves access to native
|
||||
Windows performance counters, Docker Desktop and NVIDIA telemetry.
|
||||
|
||||
The normalizer exposes a read-only normalized telemetry adapter on
|
||||
`127.0.0.1:18030`. Mission Core reads this adapter; the browser never receives MQTT or
|
||||
Timescale credentials, and Timescale is not published on a host port.
|
||||
|
||||
The stack is one deployment contour, not one multi-process container. Keeping broker,
|
||||
normalizer and database in separate containers preserves independent health checks,
|
||||
least-privilege boundaries and rollback while Compose provides one operator lifecycle:
|
||||
|
||||
```bash
|
||||
docker compose ps
|
||||
docker compose up -d
|
||||
docker compose down
|
||||
```
|
||||
|
||||
All NODE.DC-owned Docker objects use the lowercase `ndc-` namespace. Docker names are
|
||||
case-sensitive identifiers, so the product prefix is normalized to lowercase while the
|
||||
product name remains NODE.DC in operator-facing copy.
|
||||
|
||||
This directory does not contain credentials. Copy `.env.example` to `.env`, generate
|
||||
unique passwords and build the ACL/password files before starting the stack.
|
||||
|
||||
```bash
|
||||
uv run python prepare.py --initialize --mqtt-bind-address <MISSION_CORE_HOST_LAN_IP>
|
||||
docker compose up -d --build
|
||||
```
|
||||
|
||||
Expected Docker object names:
|
||||
|
||||
```text
|
||||
project: ndc-mission-core-telemetry
|
||||
network: ndc-mission-core-telemetry
|
||||
containers:
|
||||
ndc-mission-core-mqtt-broker
|
||||
ndc-mission-core-telemetry-normalizer
|
||||
ndc-mission-core-telemetry-timescaledb
|
||||
volumes:
|
||||
ndc-mission-core-mqtt-data
|
||||
ndc-mission-core-telemetry-timescale-data
|
||||
```
|
||||
|
||||
`prepare.py` passes passwords to `mosquitto_passwd` through stdin. Secrets are not placed
|
||||
in process arguments or committed files. `--initialize` generates `.env` with mode
|
||||
`0600` and refuses to replace existing credentials. The generated `.env` and `runtime/`
|
||||
directory are ignored by Git.
|
||||
|
||||
The product architecture and topic contract are defined in
|
||||
`docs/adr/0031-local-compute-contour-telemetry-plane.md`.
|
||||
|
||||
## Worker agent
|
||||
|
||||
Worker 006 uses the official Windows Telegraf distribution as the host service
|
||||
`NDC Mission Core Telemetry Agent`. Install or update it with:
|
||||
|
||||
```powershell
|
||||
.\telegraf\Install-NdcMissionCoreTelegraf.ps1
|
||||
.\telegraf\Update-NdcMissionCoreTelegraf.ps1
|
||||
```
|
||||
|
||||
The update path validates the candidate configuration, backs up the active
|
||||
configuration and rolls back if the service does not return to `Running`. MQTT
|
||||
credentials are scoped to the service environment and must not be passed on a command
|
||||
line or stored in the repository.
|
||||
|
||||
The stack and agent are intentionally not started by repository tests. Provisioning a
|
||||
machine is a separate, explicit operation.
|
||||
@@ -0,0 +1,112 @@
|
||||
name: ndc-mission-core-telemetry
|
||||
|
||||
x-ndc-labels: &ndc-labels
|
||||
com.nodedc.product: mission-core
|
||||
com.nodedc.stack: ndc-mission-core-telemetry
|
||||
com.nodedc.managed-by: docker-compose
|
||||
|
||||
services:
|
||||
broker:
|
||||
image: eclipse-mosquitto:2.1.2-alpine
|
||||
container_name: ndc-mission-core-mqtt-broker
|
||||
restart: unless-stopped
|
||||
labels:
|
||||
<<: *ndc-labels
|
||||
com.nodedc.role: mqtt-broker
|
||||
ports:
|
||||
- "${MISSIONCORE_MQTT_BIND_ADDRESS:-127.0.0.1}:${MISSIONCORE_MQTT_PORT:-1883}:1883"
|
||||
environment:
|
||||
MISSIONCORE_MQTT_HEALTH_USER: "${MISSIONCORE_MQTT_INGEST_USER}"
|
||||
MISSIONCORE_MQTT_HEALTH_PASSWORD: "${MISSIONCORE_MQTT_INGEST_PASSWORD}"
|
||||
volumes:
|
||||
- ./mosquitto/mosquitto.conf:/mosquitto/config/mosquitto.conf:ro
|
||||
- ./runtime/mosquitto/passwords:/mosquitto/config/passwords:ro
|
||||
- ./runtime/mosquitto/acl:/mosquitto/config/acl:ro
|
||||
- broker-data:/mosquitto/data
|
||||
healthcheck:
|
||||
test:
|
||||
- CMD-SHELL
|
||||
- >-
|
||||
mosquitto_sub -h 127.0.0.1 -t '$$SYS/broker/uptime' -C 1 -W 3
|
||||
-u "$$MISSIONCORE_MQTT_HEALTH_USER"
|
||||
-P "$$MISSIONCORE_MQTT_HEALTH_PASSWORD"
|
||||
interval: 10s
|
||||
timeout: 4s
|
||||
retries: 6
|
||||
|
||||
timescale:
|
||||
image: timescale/timescaledb-ha:pg16.14-ts2.28.2-all-oss
|
||||
container_name: ndc-mission-core-telemetry-timescaledb
|
||||
restart: unless-stopped
|
||||
labels:
|
||||
<<: *ndc-labels
|
||||
com.nodedc.role: telemetry-database
|
||||
environment:
|
||||
POSTGRES_DB: "${MISSIONCORE_DB_NAME}"
|
||||
POSTGRES_USER: "${MISSIONCORE_DB_USER}"
|
||||
POSTGRES_PASSWORD: "${MISSIONCORE_DB_PASSWORD}"
|
||||
volumes:
|
||||
- timescale-data:/home/postgres/pgdata/data
|
||||
- ./timescale/001_telemetry.sql:/docker-entrypoint-initdb.d/001_telemetry.sql:ro
|
||||
healthcheck:
|
||||
test: ["CMD-SHELL", "pg_isready -U ${MISSIONCORE_DB_USER} -d ${MISSIONCORE_DB_NAME}"]
|
||||
interval: 10s
|
||||
timeout: 5s
|
||||
retries: 10
|
||||
|
||||
normalizer:
|
||||
image: nodedc/mission-core-telemetry-normalizer:local
|
||||
container_name: ndc-mission-core-telemetry-normalizer
|
||||
build:
|
||||
context: ../..
|
||||
dockerfile: deploy/telemetry-plane/normalizer/Dockerfile
|
||||
restart: unless-stopped
|
||||
labels:
|
||||
<<: *ndc-labels
|
||||
com.nodedc.role: telemetry-normalizer
|
||||
depends_on:
|
||||
broker:
|
||||
condition: service_healthy
|
||||
timescale:
|
||||
condition: service_healthy
|
||||
ports:
|
||||
- "127.0.0.1:${MISSIONCORE_TELEMETRY_QUERY_PORT:-18030}:18030"
|
||||
environment:
|
||||
MISSIONCORE_MQTT_HOST: broker
|
||||
MISSIONCORE_MQTT_PORT: "1883"
|
||||
MISSIONCORE_MQTT_USERNAME: "${MISSIONCORE_MQTT_INGEST_USER}"
|
||||
MISSIONCORE_MQTT_PASSWORD: "${MISSIONCORE_MQTT_INGEST_PASSWORD}"
|
||||
MISSIONCORE_DATABASE_DSN: >-
|
||||
postgresql://${MISSIONCORE_DB_USER}:${MISSIONCORE_DB_PASSWORD}@timescale:5432/${MISSIONCORE_DB_NAME}
|
||||
MISSIONCORE_QUERY_HOST: "0.0.0.0"
|
||||
MISSIONCORE_QUERY_PORT: "18030"
|
||||
healthcheck:
|
||||
test:
|
||||
- CMD
|
||||
- python
|
||||
- -c
|
||||
- >-
|
||||
import urllib.request;
|
||||
urllib.request.urlopen('http://127.0.0.1:18030/health', timeout=3).read()
|
||||
interval: 10s
|
||||
timeout: 4s
|
||||
retries: 6
|
||||
|
||||
volumes:
|
||||
broker-data:
|
||||
name: ndc-mission-core-mqtt-data
|
||||
labels:
|
||||
<<: *ndc-labels
|
||||
com.nodedc.role: mqtt-storage
|
||||
timescale-data:
|
||||
name: ndc-mission-core-telemetry-timescale-data
|
||||
labels:
|
||||
<<: *ndc-labels
|
||||
com.nodedc.role: telemetry-storage
|
||||
|
||||
networks:
|
||||
default:
|
||||
name: ndc-mission-core-telemetry
|
||||
labels:
|
||||
<<: *ndc-labels
|
||||
com.nodedc.role: telemetry-network
|
||||
@@ -0,0 +1,11 @@
|
||||
persistence true
|
||||
persistence_location /mosquitto/data/
|
||||
|
||||
listener 1883 0.0.0.0
|
||||
allow_anonymous false
|
||||
password_file /mosquitto/config/passwords
|
||||
acl_file /mosquitto/config/acl
|
||||
|
||||
log_dest stdout
|
||||
connection_messages true
|
||||
log_timestamp true
|
||||
@@ -0,0 +1,8 @@
|
||||
FROM python:3.12-slim
|
||||
|
||||
WORKDIR /app
|
||||
RUN pip install --no-cache-dir "paho-mqtt>=2.1,<3" "psycopg[binary]>=3.2,<4"
|
||||
COPY deploy/telemetry-plane/normalizer/normalizer.py /app/normalizer.py
|
||||
|
||||
USER 65532:65532
|
||||
ENTRYPOINT ["python", "/app/normalizer.py"]
|
||||
@@ -0,0 +1,378 @@
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import os
|
||||
import re
|
||||
import threading
|
||||
import time
|
||||
from datetime import UTC, datetime
|
||||
from http import HTTPStatus
|
||||
from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
|
||||
from typing import Any, Final
|
||||
from urllib.parse import parse_qs, urlparse
|
||||
|
||||
TOPIC: Final = "mission-core/v1/contours/+/agents/+/+"
|
||||
TOPIC_RE: Final = re.compile(
|
||||
r"^mission-core/v1/contours/(?P<contour>[-a-z0-9]+)/"
|
||||
r"agents/(?P<agent>[-a-z0-9]+)/(?P<kind>host|runtime|pipeline|heartbeat)$"
|
||||
)
|
||||
SOURCE_SCHEMAS: Final = {
|
||||
"host": "missioncore.agent-host-telemetry/v1",
|
||||
"runtime": "missioncore.agent-runtime-telemetry/v1",
|
||||
"pipeline": "missioncore.agent-pipeline-telemetry/v1",
|
||||
"heartbeat": "missioncore.agent-heartbeat/v1",
|
||||
}
|
||||
TELEGRAF_SCHEMA: Final = "telegraf.metric-json/v1"
|
||||
QUERY_SCHEMA: Final = "missioncore.telemetry-query/v1"
|
||||
SAFE_IDENTIFIER: Final = re.compile(r"^[a-z0-9](?:[a-z0-9-]{0,62}[a-z0-9])?$")
|
||||
|
||||
|
||||
def _required(name: str) -> str:
|
||||
value = os.environ.get(name, "").strip()
|
||||
if not value:
|
||||
raise RuntimeError(f"{name} is required")
|
||||
return value
|
||||
|
||||
|
||||
def _timestamp(value: object) -> datetime:
|
||||
if isinstance(value, int | float) and not isinstance(value, bool):
|
||||
return datetime.fromtimestamp(float(value), tz=UTC)
|
||||
if not isinstance(value, str):
|
||||
raise ValueError("sample timestamp is required")
|
||||
parsed = datetime.fromisoformat(value.replace("Z", "+00:00"))
|
||||
if parsed.tzinfo is None:
|
||||
raise ValueError("observed_at_utc must include a timezone")
|
||||
return parsed.astimezone(UTC)
|
||||
|
||||
|
||||
def _optional_text(document: dict[str, Any], name: str) -> str | None:
|
||||
value = document.get(name)
|
||||
return value if isinstance(value, str) and value else None
|
||||
|
||||
|
||||
def _optional_integer(document: dict[str, Any], name: str) -> int | None:
|
||||
value = document.get(name)
|
||||
return value if isinstance(value, int) and not isinstance(value, bool) else None
|
||||
|
||||
|
||||
def _tag_text(document: dict[str, Any], name: str) -> str | None:
|
||||
tags = document.get("tags")
|
||||
if not isinstance(tags, dict):
|
||||
return None
|
||||
value = tags.get(name)
|
||||
return value if isinstance(value, str) and value else None
|
||||
|
||||
|
||||
def _series_key(document: dict[str, Any]) -> str:
|
||||
tags = document.get("tags")
|
||||
if not isinstance(tags, dict):
|
||||
value = document.get("series_key")
|
||||
return value if isinstance(value, str) else ""
|
||||
stable_tags = {
|
||||
str(name): str(value)
|
||||
for name, value in tags.items()
|
||||
if name not in {"agent_id", "contour_id", "host", "node_id"}
|
||||
}
|
||||
return json.dumps(
|
||||
stable_tags,
|
||||
ensure_ascii=False,
|
||||
sort_keys=True,
|
||||
separators=(",", ":"),
|
||||
)
|
||||
|
||||
|
||||
def _normalize(topic: str, payload: bytes) -> tuple[object, ...]:
|
||||
match = TOPIC_RE.fullmatch(topic)
|
||||
if match is None:
|
||||
raise ValueError("topic is outside the telemetry contract")
|
||||
document = json.loads(payload.decode("utf-8"))
|
||||
if not isinstance(document, dict):
|
||||
raise ValueError("payload must be an object")
|
||||
kind = match.group("kind")
|
||||
source_schema = document.get("schema_version")
|
||||
if source_schema == SOURCE_SCHEMAS[kind]:
|
||||
observed_value = document.get("observed_at_utc")
|
||||
measurement = kind
|
||||
elif (
|
||||
kind == "host"
|
||||
and isinstance(document.get("name"), str)
|
||||
and isinstance(document.get("fields"), dict)
|
||||
):
|
||||
source_schema = TELEGRAF_SCHEMA
|
||||
observed_value = document.get("timestamp")
|
||||
measurement = document["name"].strip()
|
||||
if not measurement:
|
||||
raise ValueError("Telegraf measurement name is required")
|
||||
else:
|
||||
raise ValueError("payload schema does not match topic kind")
|
||||
node_id = document.get("node_id") or _tag_text(document, "node_id")
|
||||
if not node_id:
|
||||
node_id = _tag_text(document, "host")
|
||||
if not isinstance(node_id, str) or not node_id.strip():
|
||||
raise ValueError("node_id is required")
|
||||
observed_at = _timestamp(observed_value)
|
||||
return (
|
||||
observed_at,
|
||||
match.group("contour"),
|
||||
match.group("agent"),
|
||||
node_id.strip(),
|
||||
kind,
|
||||
measurement,
|
||||
_series_key(document),
|
||||
source_schema,
|
||||
topic,
|
||||
_optional_text(document, "lab_id") or _tag_text(document, "lab_id"),
|
||||
_optional_text(document, "run_id") or _tag_text(document, "run_id"),
|
||||
_optional_text(document, "request_id") or _tag_text(document, "request_id"),
|
||||
_optional_integer(document, "frame_index"),
|
||||
json.dumps(document, separators=(",", ":"), ensure_ascii=False),
|
||||
)
|
||||
|
||||
|
||||
class TelemetryQueryServer(ThreadingHTTPServer):
|
||||
daemon_threads = True
|
||||
|
||||
def __init__(self, server_address: tuple[str, int], dsn: str) -> None:
|
||||
super().__init__(server_address, TelemetryQueryHandler)
|
||||
self.dsn = dsn
|
||||
|
||||
|
||||
class TelemetryQueryHandler(BaseHTTPRequestHandler):
|
||||
server: TelemetryQueryServer
|
||||
|
||||
def do_GET(self) -> None: # noqa: N802 - stdlib HTTP handler contract
|
||||
parsed = urlparse(self.path)
|
||||
if parsed.path == "/health":
|
||||
self._health()
|
||||
return
|
||||
match = re.fullmatch(
|
||||
r"/v1/contours/(?P<contour>[-a-z0-9]+)/agents/"
|
||||
r"(?P<agent>[-a-z0-9]+)/latest",
|
||||
parsed.path,
|
||||
)
|
||||
if match is None:
|
||||
self._json(HTTPStatus.NOT_FOUND, {"detail": "not found"})
|
||||
return
|
||||
query = parse_qs(parsed.query)
|
||||
try:
|
||||
max_age_seconds = int(query.get("max_age_seconds", ["120"])[0])
|
||||
except ValueError:
|
||||
self._json(HTTPStatus.BAD_REQUEST, {"detail": "invalid max_age_seconds"})
|
||||
return
|
||||
if not 1 <= max_age_seconds <= 86_400:
|
||||
self._json(HTTPStatus.BAD_REQUEST, {"detail": "invalid max_age_seconds"})
|
||||
return
|
||||
self._latest(
|
||||
match.group("contour"),
|
||||
match.group("agent"),
|
||||
max_age_seconds=max_age_seconds,
|
||||
)
|
||||
|
||||
def log_message(self, format: str, *args: object) -> None:
|
||||
print(f"telemetry query: {format % args}", flush=True)
|
||||
|
||||
def _health(self) -> None:
|
||||
import psycopg # type: ignore[import-not-found]
|
||||
|
||||
try:
|
||||
with (
|
||||
psycopg.connect(self.server.dsn) as connection,
|
||||
connection.cursor() as cursor,
|
||||
):
|
||||
cursor.execute("SELECT 1")
|
||||
cursor.fetchone()
|
||||
except psycopg.Error:
|
||||
self._json(
|
||||
HTTPStatus.SERVICE_UNAVAILABLE,
|
||||
{"ok": False, "schema_version": QUERY_SCHEMA},
|
||||
)
|
||||
return
|
||||
self._json(HTTPStatus.OK, {"ok": True, "schema_version": QUERY_SCHEMA})
|
||||
|
||||
def _latest(
|
||||
self,
|
||||
contour_id: str,
|
||||
agent_id: str,
|
||||
*,
|
||||
max_age_seconds: int,
|
||||
) -> None:
|
||||
import psycopg # type: ignore[import-not-found]
|
||||
|
||||
if (
|
||||
SAFE_IDENTIFIER.fullmatch(contour_id) is None
|
||||
or SAFE_IDENTIFIER.fullmatch(agent_id) is None
|
||||
):
|
||||
self._json(HTTPStatus.BAD_REQUEST, {"detail": "invalid identifier"})
|
||||
return
|
||||
try:
|
||||
with (
|
||||
psycopg.connect(self.server.dsn) as connection,
|
||||
connection.cursor() as cursor,
|
||||
):
|
||||
cursor.execute(
|
||||
"""
|
||||
SELECT DISTINCT ON (kind, measurement, series_key)
|
||||
observed_at, ingested_at, node_id, kind, measurement,
|
||||
series_key, source_schema, lab_id, run_id, request_id,
|
||||
frame_index, payload
|
||||
FROM contour_telemetry_samples
|
||||
WHERE contour_id = %s
|
||||
AND agent_id = %s
|
||||
AND observed_at >= NOW() - (%s * INTERVAL '1 second')
|
||||
ORDER BY kind, measurement, series_key, observed_at DESC
|
||||
""",
|
||||
(contour_id, agent_id, max_age_seconds),
|
||||
)
|
||||
rows = cursor.fetchall()
|
||||
except psycopg.Error:
|
||||
self._json(
|
||||
HTTPStatus.SERVICE_UNAVAILABLE,
|
||||
{"detail": "telemetry database unavailable"},
|
||||
)
|
||||
return
|
||||
samples = [
|
||||
{
|
||||
"observed_at_utc": row[0].astimezone(UTC).isoformat().replace("+00:00", "Z"),
|
||||
"ingested_at_utc": row[1].astimezone(UTC).isoformat().replace("+00:00", "Z"),
|
||||
"node_id": row[2],
|
||||
"kind": row[3],
|
||||
"measurement": row[4],
|
||||
"series_key": row[5],
|
||||
"source_schema": row[6],
|
||||
"lab_id": row[7],
|
||||
"run_id": row[8],
|
||||
"request_id": row[9],
|
||||
"frame_index": row[10],
|
||||
"payload": row[11],
|
||||
}
|
||||
for row in rows
|
||||
]
|
||||
self._json(
|
||||
HTTPStatus.OK,
|
||||
{
|
||||
"schema_version": QUERY_SCHEMA,
|
||||
"contour_id": contour_id,
|
||||
"agent_id": agent_id,
|
||||
"samples": samples,
|
||||
},
|
||||
)
|
||||
|
||||
def _json(self, status: HTTPStatus, document: dict[str, object]) -> None:
|
||||
body = json.dumps(
|
||||
document,
|
||||
ensure_ascii=False,
|
||||
separators=(",", ":"),
|
||||
).encode("utf-8")
|
||||
self.send_response(status.value)
|
||||
self.send_header("Content-Type", "application/json; charset=utf-8")
|
||||
self.send_header("Content-Length", str(len(body)))
|
||||
self.end_headers()
|
||||
self.wfile.write(body)
|
||||
|
||||
|
||||
def _start_query_server(dsn: str) -> ThreadingHTTPServer:
|
||||
host = os.environ.get("MISSIONCORE_QUERY_HOST", "0.0.0.0")
|
||||
port = int(os.environ.get("MISSIONCORE_QUERY_PORT", "18030"))
|
||||
server = TelemetryQueryServer((host, port), dsn)
|
||||
thread = threading.Thread(
|
||||
target=server.serve_forever,
|
||||
name="telemetry-query",
|
||||
daemon=True,
|
||||
)
|
||||
thread.start()
|
||||
return server
|
||||
|
||||
|
||||
def main() -> None:
|
||||
import paho.mqtt.client as mqtt
|
||||
import psycopg # type: ignore[import-not-found]
|
||||
from paho.mqtt import MQTTException
|
||||
from paho.mqtt.enums import CallbackAPIVersion
|
||||
|
||||
dsn = _required("MISSIONCORE_DATABASE_DSN")
|
||||
host = _required("MISSIONCORE_MQTT_HOST")
|
||||
port = int(os.environ.get("MISSIONCORE_MQTT_PORT", "1883"))
|
||||
username = _required("MISSIONCORE_MQTT_USERNAME")
|
||||
password = _required("MISSIONCORE_MQTT_PASSWORD")
|
||||
_start_query_server(dsn)
|
||||
connection = psycopg.connect(dsn, autocommit=True)
|
||||
client = mqtt.Client(
|
||||
callback_api_version=CallbackAPIVersion.VERSION2,
|
||||
client_id="ndc-mission-core-telemetry-normalizer",
|
||||
clean_session=False,
|
||||
)
|
||||
client.username_pw_set(username, password)
|
||||
|
||||
def on_connect(
|
||||
connected_client: Any,
|
||||
_userdata: object,
|
||||
_flags: Any,
|
||||
reason_code: Any,
|
||||
_properties: Any,
|
||||
) -> None:
|
||||
if reason_code.is_failure:
|
||||
raise RuntimeError(f"MQTT connection rejected: {reason_code}")
|
||||
connected_client.subscribe(TOPIC, qos=1)
|
||||
|
||||
def on_message(_client: Any, _userdata: object, message: Any) -> None:
|
||||
try:
|
||||
row = _normalize(message.topic, message.payload)
|
||||
with connection.cursor() as cursor:
|
||||
cursor.execute(
|
||||
"""
|
||||
INSERT INTO contour_telemetry_samples (
|
||||
observed_at, contour_id, agent_id, node_id, kind, measurement,
|
||||
series_key, source_schema, source_topic, lab_id, run_id,
|
||||
request_id, frame_index, payload
|
||||
)
|
||||
VALUES (
|
||||
%s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s,
|
||||
%s::jsonb
|
||||
)
|
||||
ON CONFLICT (
|
||||
observed_at, contour_id, agent_id, kind, measurement, series_key
|
||||
)
|
||||
DO UPDATE SET
|
||||
ingested_at = NOW(),
|
||||
node_id = EXCLUDED.node_id,
|
||||
series_key = EXCLUDED.series_key,
|
||||
source_schema = EXCLUDED.source_schema,
|
||||
source_topic = EXCLUDED.source_topic,
|
||||
lab_id = EXCLUDED.lab_id,
|
||||
run_id = EXCLUDED.run_id,
|
||||
request_id = EXCLUDED.request_id,
|
||||
frame_index = EXCLUDED.frame_index,
|
||||
payload = CASE
|
||||
WHEN contour_telemetry_samples.source_schema = %s
|
||||
AND EXCLUDED.source_schema = %s
|
||||
THEN EXCLUDED.payload || jsonb_build_object(
|
||||
'fields',
|
||||
COALESCE(
|
||||
contour_telemetry_samples.payload -> 'fields',
|
||||
'{}'::jsonb
|
||||
) || COALESCE(
|
||||
EXCLUDED.payload -> 'fields',
|
||||
'{}'::jsonb
|
||||
)
|
||||
)
|
||||
ELSE EXCLUDED.payload
|
||||
END
|
||||
""",
|
||||
(*row, TELEGRAF_SCHEMA, TELEGRAF_SCHEMA),
|
||||
)
|
||||
except (ValueError, UnicodeDecodeError, json.JSONDecodeError, psycopg.Error) as exc:
|
||||
print(f"telemetry sample rejected: {exc}", flush=True)
|
||||
|
||||
client.on_connect = on_connect
|
||||
client.on_message = on_message
|
||||
while True:
|
||||
try:
|
||||
client.connect(host, port, keepalive=30)
|
||||
client.loop_forever(retry_first_connection=True)
|
||||
except (OSError, MQTTException, psycopg.Error) as exc:
|
||||
print(f"telemetry normalizer reconnecting: {exc}", flush=True)
|
||||
time.sleep(3)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -0,0 +1,130 @@
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import ipaddress
|
||||
import os
|
||||
import secrets
|
||||
import subprocess
|
||||
from pathlib import Path
|
||||
from typing import Final
|
||||
|
||||
ROOT: Final = Path(__file__).resolve().parent
|
||||
ENV_PATH: Final = ROOT / ".env"
|
||||
RUNTIME: Final = ROOT / "runtime" / "mosquitto"
|
||||
IMAGE: Final = "eclipse-mosquitto:2.1.2-alpine"
|
||||
PLACEHOLDER: Final = "replace-with-"
|
||||
|
||||
|
||||
def _initialize_environment(bind_address: str, *, overwrite: bool = False) -> None:
|
||||
if ENV_PATH.exists() and not overwrite:
|
||||
raise RuntimeError(".env already exists; refusing to overwrite local credentials")
|
||||
normalized_bind_address = str(ipaddress.ip_address(bind_address))
|
||||
values = {
|
||||
"MISSIONCORE_MQTT_BIND_ADDRESS": normalized_bind_address,
|
||||
"MISSIONCORE_MQTT_PORT": "1883",
|
||||
"MISSIONCORE_TELEMETRY_QUERY_PORT": "18030",
|
||||
"MISSIONCORE_DB_NAME": "missioncore_telemetry",
|
||||
"MISSIONCORE_DB_USER": "missioncore_ingest",
|
||||
"MISSIONCORE_DB_PASSWORD": secrets.token_urlsafe(36),
|
||||
"MISSIONCORE_MQTT_INGEST_USER": "missioncore-ingest",
|
||||
"MISSIONCORE_MQTT_INGEST_PASSWORD": secrets.token_urlsafe(36),
|
||||
"MISSIONCORE_MQTT_WORKER_006_USER": "worker-006",
|
||||
"MISSIONCORE_MQTT_WORKER_006_PASSWORD": secrets.token_urlsafe(36),
|
||||
}
|
||||
ENV_PATH.write_text(
|
||||
"".join(f"{name}={value}\n" for name, value in values.items()),
|
||||
encoding="utf-8",
|
||||
)
|
||||
os.chmod(ENV_PATH, 0o600)
|
||||
|
||||
|
||||
def _environment() -> dict[str, str]:
|
||||
if not ENV_PATH.is_file():
|
||||
raise RuntimeError("copy .env.example to .env and set unique secrets first")
|
||||
values: dict[str, str] = {}
|
||||
for raw_line in ENV_PATH.read_text(encoding="utf-8").splitlines():
|
||||
line = raw_line.strip()
|
||||
if not line or line.startswith("#") or "=" not in line:
|
||||
continue
|
||||
name, value = line.split("=", 1)
|
||||
values[name.strip()] = value.strip()
|
||||
return values
|
||||
|
||||
|
||||
def _required(values: dict[str, str], name: str) -> str:
|
||||
value = values.get(name, "")
|
||||
if not value or value.startswith(PLACEHOLDER):
|
||||
raise RuntimeError(f"{name} must contain a non-placeholder value")
|
||||
return value
|
||||
|
||||
|
||||
def _password_entry(path: Path, username: str, password: str, *, create: bool) -> None:
|
||||
command = [
|
||||
"docker",
|
||||
"run",
|
||||
"--rm",
|
||||
"-i",
|
||||
"-v",
|
||||
f"{RUNTIME}:/out",
|
||||
IMAGE,
|
||||
"mosquitto_passwd",
|
||||
]
|
||||
if create:
|
||||
command.append("-c")
|
||||
command.extend([f"/out/{path.name}", username])
|
||||
subprocess.run(
|
||||
command,
|
||||
check=True,
|
||||
input=f"{password}\n{password}\n",
|
||||
text=True,
|
||||
)
|
||||
|
||||
|
||||
def main() -> None:
|
||||
parser = argparse.ArgumentParser()
|
||||
parser.add_argument(
|
||||
"--initialize",
|
||||
action="store_true",
|
||||
help="create a private .env with generated local credentials",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--mqtt-bind-address",
|
||||
default="127.0.0.1",
|
||||
help="host IP exposed to telemetry agents",
|
||||
)
|
||||
arguments = parser.parse_args()
|
||||
if arguments.initialize:
|
||||
_initialize_environment(arguments.mqtt_bind_address)
|
||||
values = _environment()
|
||||
ingest_user = _required(values, "MISSIONCORE_MQTT_INGEST_USER")
|
||||
ingest_password = _required(values, "MISSIONCORE_MQTT_INGEST_PASSWORD")
|
||||
worker_user = _required(values, "MISSIONCORE_MQTT_WORKER_006_USER")
|
||||
worker_password = _required(values, "MISSIONCORE_MQTT_WORKER_006_PASSWORD")
|
||||
_required(values, "MISSIONCORE_DB_PASSWORD")
|
||||
|
||||
RUNTIME.mkdir(parents=True, exist_ok=True)
|
||||
password_path = RUNTIME / "passwords"
|
||||
_password_entry(password_path, ingest_user, ingest_password, create=True)
|
||||
_password_entry(password_path, worker_user, worker_password, create=False)
|
||||
acl_path = RUNTIME / "acl"
|
||||
acl_path.write_text(
|
||||
"\n".join(
|
||||
[
|
||||
f"user {ingest_user}",
|
||||
"topic read mission-core/v1/contours/+/agents/+/+",
|
||||
"topic read $SYS/broker/uptime",
|
||||
"",
|
||||
"# Agent username must equal its stable agent id.",
|
||||
"pattern write mission-core/v1/contours/+/agents/%u/+",
|
||||
"",
|
||||
]
|
||||
),
|
||||
encoding="utf-8",
|
||||
)
|
||||
os.chmod(password_path, 0o600)
|
||||
os.chmod(acl_path, 0o600)
|
||||
print("Mosquitto password and ACL files prepared.", flush=True)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -0,0 +1,120 @@
|
||||
[CmdletBinding()]
|
||||
param(
|
||||
[string]$Version = "1.38.4",
|
||||
[string]$ExpectedSha256 = "6c7878ec319471ac85b82443baec2f3fa5dbcf1b6e2da5d5cd2cbb60fff2bb45",
|
||||
[string]$ConfigurationTemplate = "$PSScriptRoot\mission-core-windows.conf.tmpl"
|
||||
)
|
||||
|
||||
$ErrorActionPreference = "Stop"
|
||||
$serviceName = "telegraf"
|
||||
$installRoot = "C:\Program Files\NDC\Mission Core\Telegraf"
|
||||
$configurationRoot = "C:\ProgramData\NDC\MissionCore\telemetry-agent"
|
||||
$configurationPath = Join-Path $configurationRoot "telegraf.conf"
|
||||
$archiveUrl = "https://dl.influxdata.com/telegraf/releases/telegraf-$($Version)_windows_amd64.zip"
|
||||
$payload = [Console]::In.ReadToEnd() | ConvertFrom-Json
|
||||
|
||||
foreach ($name in @(
|
||||
"MISSIONCORE_CONTOUR_ID",
|
||||
"MISSIONCORE_AGENT_ID",
|
||||
"MISSIONCORE_NODE_ID",
|
||||
"MISSIONCORE_MQTT_HOST",
|
||||
"MISSIONCORE_MQTT_PORT",
|
||||
"MISSIONCORE_MQTT_USERNAME",
|
||||
"MISSIONCORE_MQTT_PASSWORD"
|
||||
)) {
|
||||
$value = $payload.$name
|
||||
if (-not $value) {
|
||||
throw "Provisioning payload is missing $name"
|
||||
}
|
||||
Set-Item -Path "Env:$name" -Value ([string]$value)
|
||||
}
|
||||
|
||||
if (Get-Service -Name $serviceName -ErrorAction SilentlyContinue) {
|
||||
throw "Service '$serviceName' already exists; refusing an implicit replacement"
|
||||
}
|
||||
if (-not (Test-Path -LiteralPath $ConfigurationTemplate -PathType Leaf)) {
|
||||
throw "Configuration template not found: $ConfigurationTemplate"
|
||||
}
|
||||
|
||||
$temporaryRoot = Join-Path $env:TEMP "ndc-mission-core-telegraf-$([Guid]::NewGuid().ToString('N'))"
|
||||
$archivePath = Join-Path $temporaryRoot "telegraf.zip"
|
||||
$expandedRoot = Join-Path $temporaryRoot "expanded"
|
||||
try {
|
||||
New-Item -ItemType Directory -Path $temporaryRoot, $expandedRoot -Force | Out-Null
|
||||
Invoke-WebRequest -UseBasicParsing -Uri $archiveUrl -OutFile $archivePath
|
||||
$actualSha256 = (Get-FileHash -LiteralPath $archivePath -Algorithm SHA256).Hash.ToLowerInvariant()
|
||||
if ($actualSha256 -ne $ExpectedSha256.ToLowerInvariant()) {
|
||||
throw "Telegraf archive SHA256 mismatch"
|
||||
}
|
||||
Expand-Archive -LiteralPath $archivePath -DestinationPath $expandedRoot
|
||||
$sourceExecutable = Get-ChildItem -Path $expandedRoot -Filter "telegraf.exe" -Recurse |
|
||||
Select-Object -First 1
|
||||
if (-not $sourceExecutable) {
|
||||
throw "telegraf.exe is missing from the verified archive"
|
||||
}
|
||||
|
||||
New-Item -ItemType Directory -Path $installRoot, $configurationRoot -Force | Out-Null
|
||||
Copy-Item -LiteralPath $sourceExecutable.FullName -Destination (Join-Path $installRoot "telegraf.exe")
|
||||
Copy-Item -LiteralPath $ConfigurationTemplate -Destination $configurationPath
|
||||
& icacls.exe $configurationRoot /inheritance:r /grant:r `
|
||||
"*S-1-5-18:(OI)(CI)F" "*S-1-5-32-544:(OI)(CI)F" | Out-Null
|
||||
if ($LASTEXITCODE -ne 0) {
|
||||
throw "Failed to restrict the telemetry agent configuration directory"
|
||||
}
|
||||
|
||||
$executable = Join-Path $installRoot "telegraf.exe"
|
||||
$validationOutput = Join-Path $temporaryRoot "validation.out.log"
|
||||
$validationError = Join-Path $temporaryRoot "validation.error.log"
|
||||
$validation = Start-Process -FilePath $executable `
|
||||
-ArgumentList @("--config", $configurationPath, "--test") `
|
||||
-NoNewWindow -Wait -PassThru `
|
||||
-RedirectStandardOutput $validationOutput `
|
||||
-RedirectStandardError $validationError
|
||||
if ($validation.ExitCode -ne 0) {
|
||||
$validationDetail = Get-Content -LiteralPath $validationError -Tail 8 |
|
||||
Out-String
|
||||
throw "Telegraf configuration validation failed: $validationDetail"
|
||||
}
|
||||
|
||||
& $executable --service install --config $configurationPath
|
||||
if ($LASTEXITCODE -ne 0) {
|
||||
throw "Telegraf service installation failed"
|
||||
}
|
||||
$serviceEnvironment = @(
|
||||
"MISSIONCORE_CONTOUR_ID=$($payload.MISSIONCORE_CONTOUR_ID)",
|
||||
"MISSIONCORE_AGENT_ID=$($payload.MISSIONCORE_AGENT_ID)",
|
||||
"MISSIONCORE_NODE_ID=$($payload.MISSIONCORE_NODE_ID)",
|
||||
"MISSIONCORE_MQTT_HOST=$($payload.MISSIONCORE_MQTT_HOST)",
|
||||
"MISSIONCORE_MQTT_PORT=$($payload.MISSIONCORE_MQTT_PORT)",
|
||||
"MISSIONCORE_MQTT_USERNAME=$($payload.MISSIONCORE_MQTT_USERNAME)",
|
||||
"MISSIONCORE_MQTT_PASSWORD=$($payload.MISSIONCORE_MQTT_PASSWORD)"
|
||||
)
|
||||
$serviceRegistryPath = "HKLM:\SYSTEM\CurrentControlSet\Services\$serviceName"
|
||||
Set-ItemProperty -Path $serviceRegistryPath -Name Environment `
|
||||
-Type MultiString -Value $serviceEnvironment
|
||||
Set-ItemProperty -Path $serviceRegistryPath -Name DisplayName `
|
||||
-Value "NDC Mission Core Telemetry Agent"
|
||||
& sc.exe config $serviceName DisplayName= "NDC Mission Core Telemetry Agent" | Out-Null
|
||||
if ($LASTEXITCODE -ne 0) {
|
||||
throw "Failed to apply the NDC service display name"
|
||||
}
|
||||
Set-Service -Name $serviceName -StartupType Automatic
|
||||
Start-Service -Name $serviceName
|
||||
$service = Get-Service -Name $serviceName
|
||||
$service.WaitForStatus([ServiceProcess.ServiceControllerStatus]::Running, [TimeSpan]::FromSeconds(20))
|
||||
|
||||
[ordered]@{
|
||||
SchemaVersion = "missioncore.telemetry-agent-install-result/v1"
|
||||
Agent = "Telegraf"
|
||||
Version = $Version
|
||||
ServiceName = $service.Name
|
||||
DisplayName = $service.DisplayName
|
||||
Status = $service.Status.ToString()
|
||||
StartType = $service.StartType.ToString()
|
||||
Configuration = $configurationPath
|
||||
Sha256 = $actualSha256
|
||||
} | ConvertTo-Json -Compress
|
||||
}
|
||||
finally {
|
||||
Remove-Item -LiteralPath $temporaryRoot -Recurse -Force -ErrorAction SilentlyContinue
|
||||
}
|
||||
@@ -0,0 +1,75 @@
|
||||
[CmdletBinding()]
|
||||
param(
|
||||
[Parameter(Mandatory = $true)]
|
||||
[string]$ConfigurationTemplate
|
||||
)
|
||||
|
||||
$ErrorActionPreference = "Stop"
|
||||
$serviceName = "telegraf"
|
||||
$installRoot = "C:\Program Files\NDC\Mission Core\Telegraf"
|
||||
$configurationRoot = "C:\ProgramData\NDC\MissionCore\telemetry-agent"
|
||||
$configurationPath = Join-Path $configurationRoot "telegraf.conf"
|
||||
$executable = Join-Path $installRoot "telegraf.exe"
|
||||
$serviceRegistryPath = "HKLM:\SYSTEM\CurrentControlSet\Services\$serviceName"
|
||||
|
||||
$service = Get-Service -Name $serviceName -ErrorAction Stop
|
||||
if (-not (Test-Path -LiteralPath $executable -PathType Leaf)) {
|
||||
throw "NDC Mission Core Telegraf executable is missing"
|
||||
}
|
||||
if (-not (Test-Path -LiteralPath $ConfigurationTemplate -PathType Leaf)) {
|
||||
throw "Configuration template not found: $ConfigurationTemplate"
|
||||
}
|
||||
|
||||
foreach ($entry in @((Get-ItemProperty -Path $serviceRegistryPath).Environment)) {
|
||||
$name, $value = $entry -split "=", 2
|
||||
if ($name -and $value) {
|
||||
Set-Item -Path "Env:$name" -Value $value
|
||||
}
|
||||
}
|
||||
|
||||
$temporaryRoot = Join-Path $env:TEMP "ndc-mission-core-telegraf-update-$([Guid]::NewGuid().ToString('N'))"
|
||||
$validationOutput = Join-Path $temporaryRoot "validation.out.log"
|
||||
$validationError = Join-Path $temporaryRoot "validation.error.log"
|
||||
$backupRoot = Join-Path $configurationRoot "backups"
|
||||
$backupPath = Join-Path $backupRoot "telegraf-$([DateTime]::UtcNow.ToString('yyyyMMddTHHmmssZ')).conf"
|
||||
try {
|
||||
New-Item -ItemType Directory -Path $temporaryRoot, $backupRoot -Force | Out-Null
|
||||
$validation = Start-Process -FilePath $executable `
|
||||
-ArgumentList @("--config", $ConfigurationTemplate, "--test") `
|
||||
-NoNewWindow -Wait -PassThru `
|
||||
-RedirectStandardOutput $validationOutput `
|
||||
-RedirectStandardError $validationError
|
||||
if ($validation.ExitCode -ne 0) {
|
||||
$validationDetail = Get-Content -LiteralPath $validationError -Tail 8 |
|
||||
Out-String
|
||||
throw "Telegraf configuration validation failed: $validationDetail"
|
||||
}
|
||||
|
||||
Copy-Item -LiteralPath $configurationPath -Destination $backupPath
|
||||
Stop-Service -Name $serviceName
|
||||
try {
|
||||
Copy-Item -LiteralPath $ConfigurationTemplate -Destination $configurationPath
|
||||
Start-Service -Name $serviceName
|
||||
$service = Get-Service -Name $serviceName
|
||||
$service.WaitForStatus(
|
||||
[ServiceProcess.ServiceControllerStatus]::Running,
|
||||
[TimeSpan]::FromSeconds(20)
|
||||
)
|
||||
}
|
||||
catch {
|
||||
Copy-Item -LiteralPath $backupPath -Destination $configurationPath
|
||||
Start-Service -Name $serviceName
|
||||
throw
|
||||
}
|
||||
|
||||
[ordered]@{
|
||||
SchemaVersion = "missioncore.telemetry-agent-config-result/v1"
|
||||
ServiceName = $serviceName
|
||||
Status = (Get-Service -Name $serviceName).Status.ToString()
|
||||
Configuration = $configurationPath
|
||||
Backup = $backupPath
|
||||
} | ConvertTo-Json -Compress
|
||||
}
|
||||
finally {
|
||||
Remove-Item -LiteralPath $temporaryRoot -Recurse -Force -ErrorAction SilentlyContinue
|
||||
}
|
||||
@@ -0,0 +1,35 @@
|
||||
[agent]
|
||||
interval = "2s"
|
||||
round_interval = true
|
||||
omit_hostname = false
|
||||
|
||||
[global_tags]
|
||||
node_id = "${MISSIONCORE_NODE_ID}"
|
||||
contour_id = "${MISSIONCORE_CONTOUR_ID}"
|
||||
agent_id = "${MISSIONCORE_AGENT_ID}"
|
||||
|
||||
[[inputs.cpu]]
|
||||
percpu = false
|
||||
totalcpu = true
|
||||
report_active = true
|
||||
|
||||
[[inputs.mem]]
|
||||
[[inputs.disk]]
|
||||
ignore_fs = ["tmpfs", "devtmpfs", "devfs"]
|
||||
[[inputs.net]]
|
||||
[[inputs.system]]
|
||||
[[inputs.nvidia_smi]]
|
||||
|
||||
[[inputs.docker]]
|
||||
endpoint = "unix:///var/run/docker.sock"
|
||||
container_name_include = []
|
||||
total = true
|
||||
|
||||
[[outputs.mqtt]]
|
||||
servers = ["tcp://${MISSIONCORE_MQTT_HOST}:${MISSIONCORE_MQTT_PORT}"]
|
||||
topic = "mission-core/v1/contours/${MISSIONCORE_CONTOUR_ID}/agents/${MISSIONCORE_AGENT_ID}/host"
|
||||
username = "${MISSIONCORE_MQTT_USERNAME}"
|
||||
password = "${MISSIONCORE_MQTT_PASSWORD}"
|
||||
qos = 1
|
||||
data_format = "json"
|
||||
json_timestamp_units = "1s"
|
||||
@@ -0,0 +1,57 @@
|
||||
[agent]
|
||||
interval = "2s"
|
||||
round_interval = true
|
||||
omit_hostname = false
|
||||
|
||||
[global_tags]
|
||||
node_id = "${MISSIONCORE_NODE_ID}"
|
||||
contour_id = "${MISSIONCORE_CONTOUR_ID}"
|
||||
agent_id = "${MISSIONCORE_AGENT_ID}"
|
||||
|
||||
[[inputs.cpu]]
|
||||
percpu = false
|
||||
totalcpu = true
|
||||
report_active = true
|
||||
|
||||
[[inputs.mem]]
|
||||
[[inputs.system]]
|
||||
[[inputs.disk]]
|
||||
ignore_fs = ["tmpfs", "devtmpfs", "devfs"]
|
||||
[[inputs.net]]
|
||||
[[inputs.win_perf_counters]]
|
||||
[[inputs.win_perf_counters.object]]
|
||||
ObjectName = "System"
|
||||
Counters = ["System Up Time"]
|
||||
Instances = ["------"]
|
||||
Measurement = "win_system"
|
||||
|
||||
[[inputs.nvidia_smi]]
|
||||
|
||||
[[inputs.docker]]
|
||||
endpoint = "npipe:////./pipe/docker_engine"
|
||||
container_name_include = []
|
||||
|
||||
[[inputs.http_response]]
|
||||
urls = ["http://127.0.0.1:8000/v2/health/ready"]
|
||||
response_timeout = "2s"
|
||||
response_status_code = 200
|
||||
name_override = "missioncore_triton_health"
|
||||
|
||||
[[inputs.prometheus]]
|
||||
urls = ["http://127.0.0.1:8002/metrics"]
|
||||
metric_version = 1
|
||||
response_timeout = "2s"
|
||||
namepass = [
|
||||
"nv_inference_request_success",
|
||||
"nv_inference_request_failure",
|
||||
"nv_inference_count",
|
||||
]
|
||||
|
||||
[[outputs.mqtt]]
|
||||
servers = ["tcp://${MISSIONCORE_MQTT_HOST}:${MISSIONCORE_MQTT_PORT}"]
|
||||
topic = "mission-core/v1/contours/${MISSIONCORE_CONTOUR_ID}/agents/${MISSIONCORE_AGENT_ID}/host"
|
||||
username = "${MISSIONCORE_MQTT_USERNAME}"
|
||||
password = "${MISSIONCORE_MQTT_PASSWORD}"
|
||||
qos = 1
|
||||
data_format = "json"
|
||||
json_timestamp_units = "1s"
|
||||
@@ -0,0 +1,43 @@
|
||||
CREATE EXTENSION IF NOT EXISTS timescaledb;
|
||||
|
||||
CREATE TABLE IF NOT EXISTS contour_telemetry_samples (
|
||||
observed_at TIMESTAMPTZ NOT NULL,
|
||||
ingested_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
|
||||
contour_id TEXT NOT NULL,
|
||||
agent_id TEXT NOT NULL,
|
||||
node_id TEXT NOT NULL,
|
||||
kind TEXT NOT NULL,
|
||||
measurement TEXT NOT NULL,
|
||||
series_key TEXT NOT NULL DEFAULT '',
|
||||
source_schema TEXT NOT NULL,
|
||||
source_topic TEXT NOT NULL,
|
||||
lab_id TEXT,
|
||||
run_id TEXT,
|
||||
request_id TEXT,
|
||||
frame_index BIGINT,
|
||||
payload JSONB NOT NULL,
|
||||
PRIMARY KEY (
|
||||
observed_at,
|
||||
contour_id,
|
||||
agent_id,
|
||||
kind,
|
||||
measurement,
|
||||
series_key
|
||||
)
|
||||
);
|
||||
|
||||
SELECT create_hypertable(
|
||||
'contour_telemetry_samples',
|
||||
by_range('observed_at'),
|
||||
if_not_exists => TRUE
|
||||
);
|
||||
|
||||
CREATE INDEX IF NOT EXISTS contour_telemetry_latest_idx
|
||||
ON contour_telemetry_samples (
|
||||
contour_id,
|
||||
agent_id,
|
||||
kind,
|
||||
measurement,
|
||||
series_key,
|
||||
observed_at DESC
|
||||
);
|
||||
Reference in New Issue
Block a user