feat(governance): close the 6 data-disease gaps — DQ monitoring, ownership, access posture, lineage & observability
Adds native Command Center features (no new containers) integrated as sub-tabs in the existing Data Explorer and Data Quality views: - Continuous Data Quality (dq_monitor.py): live completeness/uniqueness/validity/ freshness scorecards via Trino with rolling trends → DataQuality "Live Monitoring". - Ownership & stewardship (catalog_governance.py): owner/steward/tier matrix, orphan detection, business glossary; local store best-effort synced to OpenMetadata (owner PATCH) → Data Explorer "Ownership". - Access & policy posture: per-dataset compliance combining PII masking, ownership, live DQ and observability alerts vs data contracts → Data Explorer "Access & Policies". - Lineage (lineage.py): staged source→CDC→Spark→S3→Iceberg→Trino→serving graph with live row counts and column-level PII/masking tracing → Data Explorer "Lineage". - Observability (observability.py): volume/freshness/schema-drift monitoring with alerts → Data Explorer "Observability". - Shared lake_meta.py dataset registry + bounded Trino client; fast native row-count and PK-indexed freshness so monitors stay cheap on 25-54M-row tables. - LLM context (lab_context.py) enriched with DQ scores, ownership and active alerts.
This commit is contained in:
mo
+1
-1
@@ -4,7 +4,7 @@ WORKDIR /app
|
||||
RUN apt-get update && apt-get install -y --no-install-recommends curl && rm -rf /var/lib/apt/lists/*
|
||||
COPY requirements.txt .
|
||||
RUN pip install --no-cache-dir -r requirements.txt
|
||||
COPY main.py lab_context.py agent_terminal.py workload.py node_registry.py node_ops.py topology_views.py supervisor.py approval_service.py db.py dockhand_envs.py presentation.py database_inventory.py presentation_upload.py presentation_static.py storage_s3.py elasticsearch_api.py sql_console.py hdfs_api.py ssh_terminal.py pipeline_ops.py hadoop_analytics.py agent_ops.py cdc_consumer.py movements.py dataflow.py streaming_ops.py spark_workbench.py hadoop_sql.py hdfs_kafka.py webhdfs_util.py pii_catalog.py platform_context.py trino_federated.py etl_offload.py agent_activity.py hive_bench_seed.json .
|
||||
COPY main.py lab_context.py agent_terminal.py workload.py node_registry.py node_ops.py topology_views.py supervisor.py approval_service.py db.py dockhand_envs.py presentation.py database_inventory.py presentation_upload.py presentation_static.py storage_s3.py elasticsearch_api.py sql_console.py hdfs_api.py ssh_terminal.py pipeline_ops.py hadoop_analytics.py agent_ops.py cdc_consumer.py movements.py dataflow.py streaming_ops.py spark_workbench.py hadoop_sql.py hdfs_kafka.py webhdfs_util.py pii_catalog.py platform_context.py trino_federated.py etl_offload.py agent_activity.py lake_meta.py lineage.py dq_monitor.py observability.py catalog_governance.py hive_bench_seed.json .
|
||||
RUN mkdir -p /data
|
||||
ENV DATABASE_URL=sqlite:////data/atc-agents.db
|
||||
EXPOSE 3201
|
||||
|
||||
@@ -0,0 +1,314 @@
|
||||
"""Data governance: ownership, stewardship, glossary & posture
|
||||
(Diseases #3 Ownership and #5 Governance).
|
||||
|
||||
Ownership/steward assignments are kept in a local store (so the feature always
|
||||
works for the demo regardless of OpenMetadata ingestion state) and are
|
||||
best-effort mirrored to OpenMetadata. Users/teams for the assignment dropdowns
|
||||
are read from OpenMetadata when reachable. The governance *posture* view
|
||||
combines, per dataset: ownership, PII/masking (pii_catalog), live data-quality
|
||||
score (dq_monitor), observability alerts and a simple data-contract check.
|
||||
|
||||
Endpoints:
|
||||
GET /api/governance/datasets -> ownership matrix (+ orphan flag)
|
||||
GET /api/governance/users -> assignable users/teams
|
||||
POST /api/governance/assign -> set owner/steward/team/tier (native + OM)
|
||||
GET /api/governance/glossary -> business glossary terms
|
||||
GET /api/governance/posture -> combined governance posture per dataset
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
import os
|
||||
import time
|
||||
from pathlib import Path
|
||||
from typing import Any
|
||||
|
||||
import httpx
|
||||
from fastapi import APIRouter
|
||||
from fastapi.responses import JSONResponse
|
||||
from pydantic import BaseModel
|
||||
|
||||
from lake_meta import DATASETS, DATASET_BY_KEY
|
||||
|
||||
router = APIRouter(prefix="/api/governance", tags=["governance"])
|
||||
|
||||
OPENMETADATA_URL = os.getenv("OPENMETADATA_URL", "").rstrip("/")
|
||||
OPENMETADATA_TOKEN = os.getenv("OPENMETADATA_TOKEN", "")
|
||||
OWNERS_PATH = Path(os.getenv("GOVERNANCE_OWNERS_PATH", "/data/governance_owners.json"))
|
||||
CONTRACTS_PATH = Path(os.getenv("GOVERNANCE_CONTRACTS_PATH", "/data/governance_contracts.json"))
|
||||
|
||||
# Default data contracts (quality SLAs) per dataset — used when none stored.
|
||||
DEFAULT_CONTRACTS = {
|
||||
"_default": {"min_score": 80, "min_completeness": 95, "min_freshness_min": 60,
|
||||
"max_critical_alerts": 0},
|
||||
}
|
||||
|
||||
# Seed business glossary so the term list is never empty even before OM ingestion.
|
||||
SEED_GLOSSARY = [
|
||||
{"name": "Customer", "description": "A person or organization that places sales orders.",
|
||||
"related": ["customer_id", "customer_name", "customer_email"], "domain": "Sales"},
|
||||
{"name": "Order", "description": "A sales transaction with an amount, channel and status.",
|
||||
"related": ["order_id", "amount", "order_status"], "domain": "Sales"},
|
||||
{"name": "Revenue", "description": "Sum of order amounts over a period.",
|
||||
"related": ["amount", "currency"], "domain": "Sales"},
|
||||
{"name": "Employee", "description": "A member of the workforce tracked via HR events.",
|
||||
"related": ["employee_id", "department", "role_name"], "domain": "People"},
|
||||
{"name": "PII", "description": "Personally Identifiable Information — masked per policy.",
|
||||
"related": ["customer_email", "national_id", "billing_iban"], "domain": "Governance"},
|
||||
{"name": "Telemetry", "description": "Device metric readings over time.",
|
||||
"related": ["device_id", "metric_type", "metric_value"], "domain": "IoT"},
|
||||
]
|
||||
|
||||
|
||||
def _headers() -> dict[str, str]:
|
||||
h = {"Accept": "application/json"}
|
||||
if OPENMETADATA_TOKEN:
|
||||
h["Authorization"] = f"Bearer {OPENMETADATA_TOKEN}"
|
||||
return h
|
||||
|
||||
|
||||
def _load(path: Path) -> dict[str, Any]:
|
||||
try:
|
||||
return json.loads(path.read_text())
|
||||
except Exception:
|
||||
return {}
|
||||
|
||||
|
||||
def _save(path: Path, data: dict[str, Any]) -> None:
|
||||
try:
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
path.write_text(json.dumps(data, indent=2))
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
|
||||
# ── OpenMetadata best-effort ────────────────────────────────────────────────
|
||||
def _om_users() -> list[dict[str, Any]]:
|
||||
if not OPENMETADATA_URL:
|
||||
return []
|
||||
out: list[dict[str, Any]] = []
|
||||
try:
|
||||
with httpx.Client(timeout=8.0, verify=False) as c:
|
||||
r = c.get(f"{OPENMETADATA_URL}/api/v1/users?limit=50&isBot=false", headers=_headers())
|
||||
if r.status_code == 200:
|
||||
for u in r.json().get("data", []):
|
||||
out.append({"id": u.get("id"), "name": u.get("name"),
|
||||
"display": u.get("displayName") or u.get("name"), "type": "user"})
|
||||
rt = c.get(f"{OPENMETADATA_URL}/api/v1/teams?limit=50", headers=_headers())
|
||||
if rt.status_code == 200:
|
||||
for t in rt.json().get("data", []):
|
||||
out.append({"id": t.get("id"), "name": t.get("name"),
|
||||
"display": t.get("displayName") or t.get("name"), "type": "team"})
|
||||
except Exception:
|
||||
pass
|
||||
return out
|
||||
|
||||
|
||||
def _om_glossary() -> list[dict[str, Any]]:
|
||||
if not OPENMETADATA_URL:
|
||||
return []
|
||||
try:
|
||||
with httpx.Client(timeout=8.0, verify=False) as c:
|
||||
r = c.get(f"{OPENMETADATA_URL}/api/v1/glossaryTerms?limit=100", headers=_headers())
|
||||
if r.status_code == 200:
|
||||
return [{"name": t.get("name"), "description": t.get("description", ""),
|
||||
"domain": "OpenMetadata", "related": []}
|
||||
for t in r.json().get("data", [])]
|
||||
except Exception:
|
||||
pass
|
||||
return []
|
||||
|
||||
|
||||
def _om_patch_owner(om_fqn: str, user: dict[str, Any]) -> bool:
|
||||
"""Best-effort: set table owner in OpenMetadata via JSON-Patch."""
|
||||
if not OPENMETADATA_URL or not om_fqn or not user.get("id"):
|
||||
return False
|
||||
try:
|
||||
with httpx.Client(timeout=8.0, verify=False) as c:
|
||||
g = c.get(f"{OPENMETADATA_URL}/api/v1/tables/name/{om_fqn}", headers=_headers())
|
||||
if g.status_code != 200:
|
||||
return False
|
||||
patch = [{"op": "add", "path": "/owners/0",
|
||||
"value": {"id": user["id"], "type": user.get("type", "user")}}]
|
||||
h = {**_headers(), "Content-Type": "application/json-patch+json"}
|
||||
p = c.patch(f"{OPENMETADATA_URL}/api/v1/tables/name/{om_fqn}",
|
||||
headers=h, content=json.dumps(patch))
|
||||
return p.status_code in (200, 201)
|
||||
except Exception:
|
||||
return False
|
||||
|
||||
|
||||
# ── helpers pulling from sibling modules (all best-effort) ──────────────────
|
||||
def _pii_summary(pii_key: str) -> dict[str, Any]:
|
||||
try:
|
||||
from pii_catalog import get_pii
|
||||
for d in get_pii().get("datasets", []):
|
||||
if d.get("key") == pii_key:
|
||||
return {"pii_count": d.get("pii_count", 0),
|
||||
"all_masked": d.get("all_masked", False),
|
||||
"masked": sum(1 for c in d.get("pii_columns", []) if c.get("masked")),
|
||||
"unmasked": sum(1 for c in d.get("pii_columns", []) if not c.get("masked"))}
|
||||
except Exception:
|
||||
pass
|
||||
return {"pii_count": 0, "all_masked": False, "masked": 0, "unmasked": 0}
|
||||
|
||||
|
||||
def _dq_card(key: str) -> dict[str, Any]:
|
||||
try:
|
||||
from dq_monitor import scorecards_view
|
||||
for c in scorecards_view().get("cards", []):
|
||||
if c.get("key") == key:
|
||||
return {"score": c.get("score"), "issues": c.get("issues", []),
|
||||
"freshness_age_min": c.get("freshness_age_min")}
|
||||
except Exception:
|
||||
pass
|
||||
return {"score": None, "issues": []}
|
||||
|
||||
|
||||
def _obs_alerts(key: str) -> list[dict[str, Any]]:
|
||||
try:
|
||||
from observability import _state, _lock
|
||||
with _lock:
|
||||
return [{"type": a["type"], "severity": a["severity"], "message": a["message"]}
|
||||
for a in _state["alerts_active"].values() if a["dataset"] == key]
|
||||
except Exception:
|
||||
return []
|
||||
|
||||
|
||||
# ── views ───────────────────────────────────────────────────────────────────
|
||||
def datasets_view() -> dict[str, Any]:
|
||||
owners = _load(OWNERS_PATH)
|
||||
rows = []
|
||||
orphans = 0
|
||||
stewarded = 0
|
||||
for ds in DATASETS:
|
||||
rec = owners.get(ds["key"], {})
|
||||
owner = rec.get("owner")
|
||||
steward = rec.get("steward")
|
||||
if not owner:
|
||||
orphans += 1
|
||||
if steward:
|
||||
stewarded += 1
|
||||
rows.append({
|
||||
"key": ds["key"], "label": ds["label"], "engine": ds["engine"], "color": ds["color"],
|
||||
"table": ds["fqtn"], "domain": rec.get("domain") or ds.get("domain"),
|
||||
"owner": owner, "steward": steward, "team": rec.get("team"),
|
||||
"tier": rec.get("tier"), "classification": rec.get("classification"),
|
||||
"updated_at": rec.get("updated_at"),
|
||||
"orphan": not owner,
|
||||
"pii": _pii_summary(ds.get("pii_key", ds["key"])),
|
||||
})
|
||||
return {"ok": True, "datasets": rows,
|
||||
"summary": {"total": len(rows), "orphans": orphans, "stewarded": stewarded,
|
||||
"owned": len(rows) - orphans},
|
||||
"om_connected": bool(OPENMETADATA_URL)}
|
||||
|
||||
|
||||
class AssignRequest(BaseModel):
|
||||
key: str
|
||||
owner: str | None = None
|
||||
steward: str | None = None
|
||||
team: str | None = None
|
||||
tier: str | None = None
|
||||
classification: str | None = None
|
||||
|
||||
|
||||
def posture_view() -> dict[str, Any]:
|
||||
owners = _load(OWNERS_PATH)
|
||||
contracts = _load(CONTRACTS_PATH)
|
||||
default_c = DEFAULT_CONTRACTS["_default"]
|
||||
out = []
|
||||
compliant = 0
|
||||
for ds in DATASETS:
|
||||
key = ds["key"]
|
||||
rec = owners.get(key, {})
|
||||
contract = {**default_c, **(contracts.get(key, {}))}
|
||||
pii = _pii_summary(ds.get("pii_key", key))
|
||||
dq = _dq_card(key)
|
||||
alerts = _obs_alerts(key)
|
||||
crit = sum(1 for a in alerts if a["severity"] == "critical")
|
||||
checks = []
|
||||
score = dq.get("score")
|
||||
checks.append({"name": "DQ score", "ok": score is not None and score >= contract["min_score"],
|
||||
"value": score, "target": contract["min_score"]})
|
||||
checks.append({"name": "Owner assigned", "ok": bool(rec.get("owner")),
|
||||
"value": rec.get("owner") or "—", "target": "assigned"})
|
||||
checks.append({"name": "PII masked", "ok": pii["unmasked"] == 0,
|
||||
"value": f'{pii["masked"]}/{pii["pii_count"]}', "target": "all"})
|
||||
checks.append({"name": "Critical alerts", "ok": crit <= contract["max_critical_alerts"],
|
||||
"value": crit, "target": contract["max_critical_alerts"]})
|
||||
ok = all(c["ok"] for c in checks)
|
||||
if ok:
|
||||
compliant += 1
|
||||
out.append({
|
||||
"key": key, "label": ds["label"], "engine": ds["engine"], "color": ds["color"],
|
||||
"table": ds["fqtn"], "owner": rec.get("owner"), "steward": rec.get("steward"),
|
||||
"tier": rec.get("tier"), "pii": pii, "dq_score": score, "issues": dq.get("issues", []),
|
||||
"alerts": alerts, "contract": contract, "checks": checks, "compliant": ok,
|
||||
})
|
||||
return {"ok": True, "datasets": out,
|
||||
"summary": {"total": len(out), "compliant": compliant,
|
||||
"non_compliant": len(out) - compliant}}
|
||||
|
||||
|
||||
def summary_for_llm() -> dict[str, Any]:
|
||||
owners = _load(OWNERS_PATH)
|
||||
return {
|
||||
"owners": {k: {"owner": v.get("owner"), "steward": v.get("steward"), "tier": v.get("tier")}
|
||||
for k, v in owners.items()},
|
||||
"orphan_datasets": [d["key"] for d in DATASETS if not owners.get(d["key"], {}).get("owner")],
|
||||
}
|
||||
|
||||
|
||||
@router.get("/datasets")
|
||||
async def get_datasets() -> JSONResponse:
|
||||
return JSONResponse(datasets_view())
|
||||
|
||||
|
||||
@router.get("/users")
|
||||
async def get_users() -> JSONResponse:
|
||||
users = _om_users()
|
||||
if not users:
|
||||
users = [{"id": None, "name": n, "display": n, "type": "user"}
|
||||
for n in ("admin", "bart", "mo")] + \
|
||||
[{"id": None, "name": "Organization", "display": "Organization", "type": "team"}]
|
||||
return JSONResponse({"ok": True, "users": users, "om_connected": bool(OPENMETADATA_URL)})
|
||||
|
||||
|
||||
@router.post("/assign")
|
||||
async def assign(body: AssignRequest) -> JSONResponse:
|
||||
if body.key not in DATASET_BY_KEY:
|
||||
return JSONResponse({"ok": False, "error": f"unknown dataset {body.key}"}, status_code=400)
|
||||
owners = _load(OWNERS_PATH)
|
||||
rec = dict(owners.get(body.key, {}))
|
||||
for field in ("owner", "steward", "team", "tier", "classification"):
|
||||
val = getattr(body, field)
|
||||
if val is not None:
|
||||
rec[field] = val or None
|
||||
rec["updated_at"] = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())
|
||||
owners[body.key] = rec
|
||||
_save(OWNERS_PATH, owners)
|
||||
|
||||
om_synced = False
|
||||
if body.owner:
|
||||
ds = DATASET_BY_KEY[body.key]
|
||||
user = next((u for u in _om_users() if u.get("display") == body.owner or u.get("name") == body.owner), None)
|
||||
if user:
|
||||
om_synced = _om_patch_owner(ds.get("om_fqn", ""), user)
|
||||
return JSONResponse({"ok": True, "key": body.key, "record": rec, "om_synced": om_synced})
|
||||
|
||||
|
||||
@router.get("/glossary")
|
||||
async def get_glossary() -> JSONResponse:
|
||||
terms = _om_glossary()
|
||||
source = "openmetadata"
|
||||
if not terms:
|
||||
terms = SEED_GLOSSARY
|
||||
source = "seed"
|
||||
return JSONResponse({"ok": True, "source": source, "count": len(terms), "terms": terms})
|
||||
|
||||
|
||||
@router.get("/posture")
|
||||
async def get_posture() -> JSONResponse:
|
||||
return JSONResponse(posture_view())
|
||||
@@ -0,0 +1,345 @@
|
||||
"""Continuous data-quality monitoring (Disease #2 — Dirty Data).
|
||||
|
||||
Where dq-api runs Great-Expectations/Soda on *uploaded* files, this agent runs
|
||||
quality checks continuously against the *live* business tables through Trino and
|
||||
produces a per-dataset scorecard (0-100) across five dimensions:
|
||||
|
||||
completeness – non-null ratio across columns
|
||||
uniqueness – distinct/key ratio (duplicate detection)
|
||||
validity – domain rules (non-negative amounts, present timestamps)
|
||||
freshness – age of the newest record vs a threshold
|
||||
volume – row count + delta vs the previous cycle
|
||||
|
||||
A rolling score history powers trend sparklines. The loop streams the exact
|
||||
SQL it runs into the Data Custodian terminal so operators can see the checks.
|
||||
|
||||
Endpoints:
|
||||
GET /api/dq/scorecards -> all datasets, dimensions, score, trend
|
||||
GET /api/dq/scorecard/{key} -> one dataset detail
|
||||
POST /api/dq/run -> run one cycle now
|
||||
POST /api/dq/config -> {enabled, interval_s, sample}
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import threading
|
||||
import time
|
||||
from collections import deque
|
||||
from datetime import datetime, timezone
|
||||
from typing import Any
|
||||
|
||||
from fastapi import APIRouter, Body
|
||||
from fastapi.responses import JSONResponse
|
||||
|
||||
from lake_meta import DATASETS, DATASET_BY_KEY, trino, trino_scalar, discover_columns
|
||||
|
||||
router = APIRouter(prefix="/api/dq", tags=["data-quality"])
|
||||
|
||||
FRESHNESS_THRESHOLD_MIN = 60.0 # newest row older than this => freshness degraded
|
||||
MAX_COLS = 24
|
||||
|
||||
_lock = threading.Lock()
|
||||
_state: dict[str, Any] = {
|
||||
"enabled": True,
|
||||
"interval_s": 180.0,
|
||||
"sample": 20000,
|
||||
"running": False,
|
||||
"cycles": 0,
|
||||
"last_cycle_ts": 0.0,
|
||||
"next_run_ts": 0.0,
|
||||
"started_at": None,
|
||||
"cards": {}, # key -> latest scorecard
|
||||
"history": {d["key"]: deque(maxlen=60) for d in DATASETS},
|
||||
"cols_cache": {}, # key -> [{name,type}]
|
||||
"feed": deque(maxlen=40),
|
||||
}
|
||||
|
||||
|
||||
def _now() -> datetime:
|
||||
return datetime.now(timezone.utc)
|
||||
|
||||
|
||||
def _term(text: str, level: str = "info", phase: str = "dq") -> None:
|
||||
try:
|
||||
from agent_terminal import emit_threadsafe
|
||||
emit_threadsafe("data-custodian", text, level=level, phase=phase)
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
|
||||
def _feed(text: str, level: str = "info") -> None:
|
||||
_state["feed"].appendleft({"ts": _now().isoformat(), "text": text, "level": level})
|
||||
|
||||
|
||||
def _score_color(score: float) -> str:
|
||||
if score >= 90:
|
||||
return "#34d399"
|
||||
if score >= 75:
|
||||
return "#fbbf24"
|
||||
if score >= 50:
|
||||
return "#fb923c"
|
||||
return "#f87171"
|
||||
|
||||
|
||||
def _obs_volume_freshness(key: str) -> tuple[int | None, float | None]:
|
||||
"""Read row count + freshness from the observability monitor (in-memory)."""
|
||||
try:
|
||||
from observability import _state, _lock
|
||||
with _lock:
|
||||
st = _state["datasets"].get(key) or {}
|
||||
return st.get("rows"), st.get("freshness_age_min")
|
||||
except Exception:
|
||||
return None, None
|
||||
|
||||
|
||||
def _columns(ds: dict[str, Any]) -> list[dict[str, str]]:
|
||||
key = ds["key"]
|
||||
cols = _state["cols_cache"].get(key)
|
||||
if not cols:
|
||||
cols = discover_columns(ds)
|
||||
if cols:
|
||||
_state["cols_cache"][key] = cols
|
||||
return cols or []
|
||||
|
||||
|
||||
def _assess(ds: dict[str, Any], sample: int) -> dict[str, Any]:
|
||||
key = ds["key"]
|
||||
fq = ds["fqtn"]
|
||||
cols = _columns(ds)
|
||||
col_names = [c["name"] for c in cols][:MAX_COLS]
|
||||
dims: dict[str, Any] = {}
|
||||
issues: list[str] = []
|
||||
col_profiles: list[dict[str, Any]] = []
|
||||
|
||||
# ── one bounded query: completeness + uniqueness + validity over a sample ──
|
||||
has_unique_key = ds.get("unique_key", True)
|
||||
key_col = ds.get("key_col") if ds.get("key_col") in col_names else (col_names[0] if col_names else None)
|
||||
if not has_unique_key:
|
||||
key_col = None # no single-column unique key (e.g. composite PK) → skip dedup
|
||||
ts_col = ds.get("ts_col") if ds.get("ts_col") in col_names else None
|
||||
amt_col = ds.get("amount_col") if ds.get("amount_col") in col_names else None
|
||||
|
||||
selects = ["count(*) AS n"]
|
||||
for i, c in enumerate(col_names):
|
||||
selects.append(f'count("{c}") AS c{i}')
|
||||
if key_col:
|
||||
selects.append(f'count(DISTINCT "{key_col}") AS dk')
|
||||
if amt_col:
|
||||
selects.append(f'count_if("{amt_col}" >= 0) AS amt_ok')
|
||||
if ts_col:
|
||||
selects.append(f'count_if("{ts_col}" IS NOT NULL) AS ts_ok')
|
||||
|
||||
sql = f"SELECT {', '.join(selects)} FROM (SELECT * FROM {fq} LIMIT {sample}) s"
|
||||
_term(f"$ trino: profile {key} ({len(col_names)} cols · sample {sample:,})", level="cmd", phase="dq")
|
||||
cols_out, rows = trino(sql, timeout=40.0)
|
||||
rec = dict(zip(cols_out, rows[0])) if rows else {}
|
||||
n = int(rec.get("n") or 0)
|
||||
|
||||
if n > 0:
|
||||
# completeness
|
||||
non_null_ratios = []
|
||||
for i, c in enumerate(col_names):
|
||||
cnt = int(rec.get(f"c{i}") or 0)
|
||||
ratio = cnt / n
|
||||
non_null_ratios.append(ratio)
|
||||
col_profiles.append({"name": c, "completeness": round(ratio * 100, 1),
|
||||
"nulls": n - cnt})
|
||||
completeness = round(100.0 * sum(non_null_ratios) / max(1, len(non_null_ratios)), 1)
|
||||
dims["completeness"] = completeness
|
||||
worst = sorted(col_profiles, key=lambda x: x["completeness"])[:3]
|
||||
for w in worst:
|
||||
if w["completeness"] < 95:
|
||||
issues.append(f'{w["name"]} {100 - w["completeness"]:.0f}% null')
|
||||
|
||||
# uniqueness / dedup
|
||||
if key_col and rec.get("dk") is not None:
|
||||
dk = int(rec["dk"])
|
||||
uniq = round(100.0 * dk / n, 1)
|
||||
dims["uniqueness"] = uniq
|
||||
dups = n - dk
|
||||
if dups > 0:
|
||||
issues.append(f"{dups:,} duplicate {key_col} in sample")
|
||||
|
||||
# validity
|
||||
valid_parts = []
|
||||
if amt_col and rec.get("amt_ok") is not None:
|
||||
valid_parts.append(int(rec["amt_ok"]) / n)
|
||||
if ts_col and rec.get("ts_ok") is not None:
|
||||
valid_parts.append(int(rec["ts_ok"]) / n)
|
||||
if valid_parts:
|
||||
validity = round(100.0 * sum(valid_parts) / len(valid_parts), 1)
|
||||
dims["validity"] = validity
|
||||
if validity < 99 and amt_col:
|
||||
issues.append(f"negative/invalid {amt_col}")
|
||||
|
||||
# ── volume + freshness: reuse the observability monitor's in-memory probes
|
||||
# (it already polls count(*) and max(ts)). We deliberately do NOT issue our
|
||||
# own count/max here — those are expensive full scans on 25-54M-row tables —
|
||||
# so a DQ cycle stays fast and the score is driven by the cheap sample. ──
|
||||
volume, fresh_age_min = _obs_volume_freshness(key)
|
||||
if volume is None:
|
||||
volume = n # sample size until observability reports the real count
|
||||
if fresh_age_min is not None:
|
||||
fresh_score = 100.0 if fresh_age_min <= FRESHNESS_THRESHOLD_MIN else max(
|
||||
0.0, 100.0 - (fresh_age_min - FRESHNESS_THRESHOLD_MIN) / 5.0)
|
||||
dims["freshness"] = round(fresh_score, 1)
|
||||
if fresh_age_min > FRESHNESS_THRESHOLD_MIN and not ds.get("curated"):
|
||||
issues.append(f"stale {fresh_age_min:.0f}m")
|
||||
|
||||
score = round(sum(dims.values()) / len(dims), 1) if dims else 0.0
|
||||
prev = _state["cards"].get(key, {})
|
||||
prev_vol = prev.get("volume")
|
||||
delta = (volume - prev_vol) if (volume is not None and prev_vol is not None) else None
|
||||
|
||||
card = {
|
||||
"key": key, "label": ds["label"], "engine": ds["engine"], "color": ds["color"],
|
||||
"table": fq, "domain": ds.get("domain"),
|
||||
"score": score, "score_color": _score_color(score),
|
||||
"dimensions": dims, "volume": volume, "volume_delta": delta,
|
||||
"freshness_age_min": round(fresh_age_min, 1) if fresh_age_min is not None else None,
|
||||
"issues": issues[:5], "columns": len(col_names),
|
||||
"worst_columns": sorted(col_profiles, key=lambda x: x["completeness"])[:5],
|
||||
"ts": _now().isoformat(),
|
||||
}
|
||||
lvl = "ok" if score >= 90 else ("warn" if score >= 60 else "err")
|
||||
_term(f" ← {key}: score {score} · " + " · ".join(f"{k} {v}" for k, v in dims.items())
|
||||
+ (f" · {len(issues)} issue(s)" if issues else ""), level=lvl, phase="dq")
|
||||
return card
|
||||
|
||||
|
||||
def run_cycle() -> dict[str, Any]:
|
||||
if _state["running"]:
|
||||
return {"ok": True, "skipped": "already running"}
|
||||
_state["running"] = True
|
||||
sample = int(_state["sample"])
|
||||
n_ok = 0
|
||||
try:
|
||||
_term(f"═══ DQ monitor cycle {_state['cycles'] + 1} — live tables via Trino ═══",
|
||||
level="info", phase="cycle")
|
||||
for ds in DATASETS:
|
||||
if not _state["enabled"]:
|
||||
break
|
||||
try:
|
||||
card = _assess(ds, sample)
|
||||
except Exception as exc:
|
||||
card = {"key": ds["key"], "label": ds["label"], "engine": ds["engine"],
|
||||
"color": ds["color"], "table": ds["fqtn"], "score": None,
|
||||
"score_color": "#64748b", "dimensions": {}, "error": str(exc)[:160],
|
||||
"issues": [f"check failed: {str(exc)[:80]}"], "ts": _now().isoformat()}
|
||||
_term(f" ✗ {ds['key']}: {str(exc)[:120]}", level="err", phase="dq")
|
||||
with _lock:
|
||||
_state["cards"][ds["key"]] = card
|
||||
if card.get("score") is not None:
|
||||
_state["history"][ds["key"]].append({"t": _now().strftime("%H:%M"), "score": card["score"]})
|
||||
n_ok += 1
|
||||
_state["cycles"] += 1
|
||||
_state["last_cycle_ts"] = time.time()
|
||||
_state["next_run_ts"] = time.time() + float(_state["interval_s"])
|
||||
avg = _overall_score()
|
||||
_feed(f"cycle {_state['cycles']} — {n_ok}/{len(DATASETS)} datasets scored · platform DQ {avg}")
|
||||
_term(f"═══ cycle {_state['cycles']} done — platform DQ score {avg} ═══", level="ok", phase="cycle")
|
||||
finally:
|
||||
_state["running"] = False
|
||||
return {"ok": True, "scored": n_ok, "platform_score": _overall_score()}
|
||||
|
||||
|
||||
def _overall_score() -> float | None:
|
||||
vals = [c["score"] for c in _state["cards"].values() if c.get("score") is not None]
|
||||
return round(sum(vals) / len(vals), 1) if vals else None
|
||||
|
||||
|
||||
def _loop() -> None:
|
||||
_state["started_at"] = _now().isoformat()
|
||||
time.sleep(20)
|
||||
while True:
|
||||
try:
|
||||
if _state["enabled"]:
|
||||
run_cycle()
|
||||
except Exception:
|
||||
pass
|
||||
time.sleep(max(30.0, float(_state["interval_s"])))
|
||||
|
||||
|
||||
threading.Thread(target=_loop, daemon=True, name="dq-monitor").start()
|
||||
|
||||
|
||||
def scorecards_view() -> dict[str, Any]:
|
||||
with _lock:
|
||||
cards = []
|
||||
for ds in DATASETS:
|
||||
c = _state["cards"].get(ds["key"])
|
||||
if c:
|
||||
cards.append({**c, "trend": list(_state["history"][ds["key"]])})
|
||||
else:
|
||||
cards.append({"key": ds["key"], "label": ds["label"], "engine": ds["engine"],
|
||||
"color": ds["color"], "table": ds["fqtn"], "score": None,
|
||||
"score_color": "#64748b", "dimensions": {}, "issues": [],
|
||||
"trend": [], "pending": True})
|
||||
# platform dimension averages
|
||||
dim_avg: dict[str, list[float]] = {}
|
||||
for c in cards:
|
||||
for k, v in (c.get("dimensions") or {}).items():
|
||||
dim_avg.setdefault(k, []).append(v)
|
||||
return {
|
||||
"ok": True,
|
||||
"enabled": _state["enabled"],
|
||||
"interval_s": _state["interval_s"],
|
||||
"sample": _state["sample"],
|
||||
"running": _state["running"],
|
||||
"cycles": _state["cycles"],
|
||||
"last_cycle_ts": _state["last_cycle_ts"],
|
||||
"next_run_ts": _state["next_run_ts"],
|
||||
"platform_score": _overall_score(),
|
||||
"dimension_averages": {k: round(sum(v) / len(v), 1) for k, v in dim_avg.items()},
|
||||
"cards": cards,
|
||||
"feed": list(_state["feed"])[:20],
|
||||
}
|
||||
|
||||
|
||||
def summary_for_llm() -> dict[str, Any]:
|
||||
with _lock:
|
||||
return {
|
||||
"platform_dq_score": _overall_score(),
|
||||
"datasets": [{"key": c["key"], "score": c.get("score"),
|
||||
"issues": c.get("issues", []), "freshness_age_min": c.get("freshness_age_min")}
|
||||
for c in _state["cards"].values()],
|
||||
}
|
||||
|
||||
|
||||
@router.get("/scorecards")
|
||||
async def get_scorecards() -> JSONResponse:
|
||||
return JSONResponse(scorecards_view())
|
||||
|
||||
|
||||
@router.get("/scorecard/{key}")
|
||||
async def get_scorecard(key: str) -> JSONResponse:
|
||||
with _lock:
|
||||
c = _state["cards"].get(key)
|
||||
if not c:
|
||||
return JSONResponse({"ok": False, "error": "unknown or not yet scored"}, status_code=404)
|
||||
return JSONResponse({"ok": True, "card": {**c, "trend": list(_state["history"].get(key, []))}})
|
||||
|
||||
|
||||
@router.post("/run")
|
||||
async def post_run() -> JSONResponse:
|
||||
from starlette.concurrency import run_in_threadpool
|
||||
res = await run_in_threadpool(run_cycle)
|
||||
return JSONResponse({**res, "scorecards": scorecards_view()})
|
||||
|
||||
|
||||
@router.post("/config")
|
||||
async def post_config(body: dict = Body(default={})) -> JSONResponse:
|
||||
if "enabled" in body:
|
||||
_state["enabled"] = bool(body["enabled"])
|
||||
if "interval_s" in body:
|
||||
try:
|
||||
_state["interval_s"] = max(30.0, min(1800.0, float(body["interval_s"])))
|
||||
except Exception:
|
||||
pass
|
||||
if "sample" in body:
|
||||
try:
|
||||
_state["sample"] = max(1000, min(200000, int(body["sample"])))
|
||||
except Exception:
|
||||
pass
|
||||
return JSONResponse({"ok": True, "enabled": _state["enabled"],
|
||||
"interval_s": _state["interval_s"], "sample": _state["sample"]})
|
||||
@@ -671,6 +671,21 @@ def collect_governance(log: TerminalLogFn | None = None) -> dict[str, Any]:
|
||||
"mysql_hr.hr.employee_events -> iceberg.curated_masked.employee_events_masked (PII masked)",
|
||||
"hdfs:/data/historical/sales_orders -> iceberg.hadoop.historical_sales_hdfs",
|
||||
]
|
||||
try:
|
||||
from dq_monitor import summary_for_llm as _dq
|
||||
out["data_quality"] = _dq()
|
||||
except Exception as exc:
|
||||
out["data_quality"] = {"error": str(exc)}
|
||||
try:
|
||||
from observability import summary_for_llm as _obs
|
||||
out["observability"] = _obs()
|
||||
except Exception as exc:
|
||||
out["observability"] = {"error": str(exc)}
|
||||
try:
|
||||
from catalog_governance import summary_for_llm as _own
|
||||
out["ownership"] = _own()
|
||||
except Exception as exc:
|
||||
out["ownership"] = {"error": str(exc)}
|
||||
return out
|
||||
|
||||
|
||||
@@ -700,6 +715,27 @@ def _section_governance(g: dict[str, Any]) -> list[str]:
|
||||
lines.append(" Lineage:")
|
||||
for ln in g.get("lineage") or []:
|
||||
lines.append(f" - {ln}")
|
||||
dq = g.get("data_quality") or {}
|
||||
if isinstance(dq, dict) and "error" not in dq:
|
||||
lines.append(f" Data quality (continuous, live tables): platform score {dq.get('platform_dq_score')}")
|
||||
for d in dq.get("datasets") or []:
|
||||
iss = f" issues: {', '.join(d['issues'][:3])}" if d.get("issues") else ""
|
||||
lines.append(f" - {d['key']}: score {d.get('score')}{iss}")
|
||||
own = g.get("ownership") or {}
|
||||
if isinstance(own, dict) and "error" not in own:
|
||||
orph = own.get("orphan_datasets") or []
|
||||
lines.append(f" Ownership: {len(own.get('owners', {}))} assigned"
|
||||
+ (f", orphans (no owner): {', '.join(orph)}" if orph else ", no orphans"))
|
||||
for k, v in (own.get("owners") or {}).items():
|
||||
if v.get("owner"):
|
||||
lines.append(f" - {k}: owner={v.get('owner')} steward={v.get('steward') or '—'} tier={v.get('tier') or '—'}")
|
||||
obs = g.get("observability") or {}
|
||||
if isinstance(obs, dict) and "error" not in obs:
|
||||
ac = obs.get("active_alerts") or {}
|
||||
lines.append(f" Observability alerts: {ac.get('total', 0)} active "
|
||||
f"(critical={ac.get('critical', 0)}, warning={ac.get('warning', 0)})")
|
||||
for a in (obs.get("alerts") or [])[:5]:
|
||||
lines.append(f" - [{a['severity']}] {a['dataset']}: {a['message']}")
|
||||
return lines
|
||||
|
||||
|
||||
|
||||
@@ -0,0 +1,147 @@
|
||||
"""Shared lakehouse metadata + Trino helpers.
|
||||
|
||||
Single source of truth for the business datasets the governance / data-quality /
|
||||
lineage / observability features operate on, plus a small synchronous Trino
|
||||
client. Imported by dq_monitor.py, observability.py, lineage.py and
|
||||
catalog_governance.py so every feature reasons about the exact same tables that
|
||||
the rest of the Command Center (pii_catalog, etl_offload, trino_federated)
|
||||
already exposes.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import os
|
||||
from typing import Any
|
||||
|
||||
import httpx
|
||||
|
||||
TRINO_URL = os.getenv("TRINO_URL", "http://10.0.21.50:8089").rstrip("/")
|
||||
TRINO_USER = os.getenv("TRINO_USER", "mo")
|
||||
|
||||
# Canonical business datasets, aligned with pii_catalog.DATASETS / etl_offload.
|
||||
# fqtn = fully-qualified Trino table name (catalog.schema.table)
|
||||
# om_fqn = OpenMetadata FQN best-effort (governance falls back to native store)
|
||||
DATASETS: list[dict[str, Any]] = [
|
||||
{"key": "orders", "label": "Sales orders", "engine": "PostgreSQL", "color": "#fbbf24",
|
||||
"catalog": "postgres_sales", "schema": "public", "table": "sales_orders",
|
||||
"fqtn": "postgres_sales.public.sales_orders", "pii_key": "postgres",
|
||||
"om_fqn": "atc_trino.postgres_sales.public.sales_orders",
|
||||
"key_col": "order_id", "ts_col": "order_ts", "amount_col": "amount",
|
||||
"native_count": ("postgres", "public.sales_orders"),
|
||||
"domain": "Sales", "topic": "atc.public.sales_orders"},
|
||||
{"key": "hr", "label": "HR events", "engine": "MySQL", "color": "#60a5fa",
|
||||
"catalog": "mysql_hr", "schema": "hr", "table": "employee_events",
|
||||
"fqtn": "mysql_hr.hr.employee_events", "pii_key": "mysql",
|
||||
"om_fqn": "atc_trino.mysql_hr.hr.employee_events",
|
||||
"key_col": "event_id", "ts_col": "event_ts", "amount_col": "salary_change",
|
||||
"native_count": ("mysql", "hr.employee_events"),
|
||||
"domain": "People", "topic": "atc.hr.employee_events"},
|
||||
{"key": "supply", "label": "Supply chain events", "engine": "MongoDB", "color": "#a78bfa",
|
||||
"catalog": "mongodb_supplychain", "schema": "supplychain", "table": "events",
|
||||
"fqtn": "mongodb_supplychain.supplychain.events", "pii_key": "mongodb",
|
||||
"om_fqn": "atc_trino.mongodb_supplychain.supplychain.events",
|
||||
"key_col": "event_id", "ts_col": "ts", "amount_col": "amount",
|
||||
"native_count": ("mongodb", "supplychain.events"),
|
||||
"domain": "Supply Chain", "topic": "atc.supplychain.events"},
|
||||
{"key": "telemetry", "label": "Device telemetry", "engine": "Cassandra", "color": "#22d3ee",
|
||||
"catalog": "cassandra_telemetry", "schema": "telemetry", "table": "device_metrics",
|
||||
"fqtn": "cassandra_telemetry.telemetry.device_metrics", "pii_key": "cassandra",
|
||||
"om_fqn": "atc_trino.cassandra_telemetry.telemetry.device_metrics",
|
||||
"key_col": "device_id", "ts_col": "metric_ts", "amount_col": "metric_value",
|
||||
"unique_key": False,
|
||||
"domain": "IoT", "topic": "atc.telemetry.device_metrics"},
|
||||
{"key": "curated", "label": "Curated masked (Iceberg)", "engine": "Iceberg / Trino", "color": "#34d399",
|
||||
"catalog": "iceberg", "schema": "curated_masked", "table": "sales_orders_masked",
|
||||
"fqtn": "iceberg.curated_masked.sales_orders_masked", "pii_key": "curated",
|
||||
"om_fqn": "atc_trino.iceberg.curated_masked.sales_orders_masked",
|
||||
"key_col": "order_id", "ts_col": "order_ts", "amount_col": "amount",
|
||||
"domain": "Sales", "curated": True, "topic": None},
|
||||
{"key": "hadoop", "label": "Historical sales (HDFS)", "engine": "Iceberg / HDFS", "color": "#f472b6",
|
||||
"catalog": "iceberg", "schema": "hadoop", "table": "historical_sales_hdfs",
|
||||
"fqtn": "iceberg.hadoop.historical_sales_hdfs", "pii_key": "hadoop",
|
||||
"om_fqn": "atc_trino.iceberg.hadoop.historical_sales_hdfs",
|
||||
"key_col": "order_id", "ts_col": "order_ts", "amount_col": "amount",
|
||||
"domain": "Sales", "curated": True, "topic": None},
|
||||
]
|
||||
|
||||
DATASET_BY_KEY = {d["key"]: d for d in DATASETS}
|
||||
|
||||
# Heuristics for picking the freshness / key column when discovering schema.
|
||||
_TS_HINTS = ("_ts", "ts", "updated_at", "created_at", "event_time", "modified", "time")
|
||||
_KEY_HINTS = ("_id", "id", "uuid", "key", "pk")
|
||||
|
||||
|
||||
def trino(sql: str, timeout: float = 25.0) -> tuple[list[str], list[list[Any]]]:
|
||||
"""Run a Trino statement, following nextUri pages. Returns (columns, rows).
|
||||
|
||||
`timeout` is both the per-request timeout AND an overall wall-clock deadline,
|
||||
so a long full-scan (e.g. count(*) on a huge Cassandra table) is aborted and
|
||||
cancelled instead of looping over nextUri pages for minutes and blocking the
|
||||
caller's thread.
|
||||
"""
|
||||
import time as _t
|
||||
cols: list[str] = []
|
||||
rows: list[list[Any]] = []
|
||||
deadline = _t.monotonic() + timeout
|
||||
with httpx.Client(timeout=min(timeout, 15.0)) as client:
|
||||
d = client.post(f"{TRINO_URL}/v1/statement", content=sql.encode(),
|
||||
headers={"X-Trino-User": TRINO_USER}).json()
|
||||
while True:
|
||||
if d.get("error"):
|
||||
raise RuntimeError(d["error"].get("message", "trino error"))
|
||||
c = d.get("columns")
|
||||
if c and not cols:
|
||||
cols = [x["name"] for x in c]
|
||||
rows += d.get("data") or []
|
||||
nxt = d.get("nextUri")
|
||||
if not nxt:
|
||||
break
|
||||
if _t.monotonic() > deadline:
|
||||
try:
|
||||
client.delete(nxt) # cancel the running query server-side
|
||||
except Exception:
|
||||
pass
|
||||
raise TimeoutError(f"trino query exceeded {timeout}s deadline")
|
||||
d = client.get(nxt).json()
|
||||
return cols, rows
|
||||
|
||||
|
||||
def trino_scalar(sql: str, timeout: float = 25.0) -> Any:
|
||||
_c, rows = trino(sql, timeout=timeout)
|
||||
if rows and rows[0]:
|
||||
return rows[0][0]
|
||||
return None
|
||||
|
||||
|
||||
def discover_columns(ds: dict[str, Any], timeout: float = 12.0) -> list[dict[str, str]]:
|
||||
"""[{name, type}] for a dataset's table via information_schema."""
|
||||
sql = (f"SELECT column_name, data_type FROM {ds['catalog']}.information_schema.columns "
|
||||
f"WHERE table_name = '{ds['table']}'")
|
||||
if ds.get("schema"):
|
||||
sql += f" AND table_schema = '{ds['schema']}'"
|
||||
try:
|
||||
_c, rows = trino(sql, timeout=timeout)
|
||||
return [{"name": r[0], "type": r[1]} for r in rows]
|
||||
except Exception:
|
||||
return []
|
||||
|
||||
|
||||
def pick_ts_col(columns: list[dict[str, str]], default: str | None = None) -> str | None:
|
||||
names = [c["name"] for c in columns]
|
||||
for c in columns:
|
||||
if "timestamp" in (c.get("type") or "").lower() or "date" in (c.get("type") or "").lower():
|
||||
return c["name"]
|
||||
for n in names:
|
||||
if any(h in n.lower() for h in _TS_HINTS):
|
||||
return n
|
||||
return default if default in names else None
|
||||
|
||||
|
||||
def pick_key_col(columns: list[dict[str, str]], default: str | None = None) -> str | None:
|
||||
names = [c["name"] for c in columns]
|
||||
if default in names:
|
||||
return default
|
||||
for n in names:
|
||||
if any(n.lower() == h or n.lower().endswith(h) for h in _KEY_HINTS):
|
||||
return n
|
||||
return names[0] if names else None
|
||||
+248
@@ -0,0 +1,248 @@
|
||||
"""Data lineage graph for the Command Center (Disease #6 — Traceability).
|
||||
|
||||
Models the real end-to-end pipeline as an explicit, staged node/edge graph:
|
||||
|
||||
source DB → Debezium/Kafka (CDC) → Spark / Kafka Connect → S3 Parquet lake
|
||||
→ Iceberg lakehouse (curated_masked / hadoop) → Trino federation
|
||||
→ serving (Elasticsearch, RAG/ChromaDB, vLLM, Command Center chat)
|
||||
|
||||
Table-backed nodes are enriched with live Trino row counts. For the sales path
|
||||
we expose column-level lineage with PII / masking status (reusing pii_catalog),
|
||||
so an operator can trace any field from source to the masked curated layer.
|
||||
OpenMetadata table lineage is layered in best-effort where the FQN resolves.
|
||||
|
||||
Endpoints:
|
||||
GET /api/lineage/graph?dataset=<key> -> nodes + edges (+ column links)
|
||||
GET /api/lineage/datasets -> selectable datasets
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import os
|
||||
import time
|
||||
from typing import Any
|
||||
|
||||
import httpx
|
||||
from fastapi import APIRouter
|
||||
from fastapi.responses import JSONResponse
|
||||
|
||||
from lake_meta import DATASETS, DATASET_BY_KEY, trino_scalar
|
||||
|
||||
router = APIRouter(prefix="/api/lineage", tags=["lineage"])
|
||||
|
||||
OPENMETADATA_URL = os.getenv("OPENMETADATA_URL", "").rstrip("/")
|
||||
OPENMETADATA_TOKEN = os.getenv("OPENMETADATA_TOKEN", "")
|
||||
|
||||
_cache: dict[str, Any] = {"ts": 0.0, "data": None}
|
||||
_TTL = 30.0
|
||||
|
||||
# Which source datasets flow through the CDC → lakehouse pipeline.
|
||||
_SOURCE_KEYS = ["orders", "hr", "supply", "telemetry"]
|
||||
|
||||
|
||||
def _is_active() -> tuple[bool, bool]:
|
||||
gen = arch = False
|
||||
try:
|
||||
from trino_federated import generator_active
|
||||
gen = bool(generator_active())
|
||||
except Exception:
|
||||
pass
|
||||
try:
|
||||
from storage_s3 import archive_active
|
||||
arch = bool(archive_active())
|
||||
except Exception:
|
||||
pass
|
||||
return gen, arch
|
||||
|
||||
|
||||
def _obs_counts() -> dict[str, int | None]:
|
||||
"""Reuse the row counts the observability monitor already polls in-memory,
|
||||
so the lineage graph never issues its own (potentially slow) count(*)."""
|
||||
try:
|
||||
from observability import _state, _lock
|
||||
with _lock:
|
||||
return {k: v.get("rows") for k, v in _state["datasets"].items()}
|
||||
except Exception:
|
||||
return {}
|
||||
|
||||
|
||||
def _row_count(ds_key: str, counts: dict[str, int | None]) -> int | None:
|
||||
return counts.get(ds_key)
|
||||
|
||||
|
||||
def _pii_columns(ds_key: str) -> list[dict[str, Any]]:
|
||||
"""Column-level detail with PII/masking flags from pii_catalog (best-effort).
|
||||
|
||||
ds_key is a lake_meta key; map it to the pii_catalog key first."""
|
||||
pii_key = (DATASET_BY_KEY.get(ds_key, {}) or {}).get("pii_key", ds_key)
|
||||
try:
|
||||
from pii_catalog import get_pii
|
||||
data = get_pii()
|
||||
for d in data.get("datasets", []):
|
||||
if d.get("key") == pii_key or d.get("node_id") == pii_key:
|
||||
return [{"name": c["name"], "category": c.get("category"),
|
||||
"masked": bool(c.get("masked")), "pii": True}
|
||||
for c in d.get("pii_columns", [])]
|
||||
except Exception:
|
||||
pass
|
||||
return []
|
||||
|
||||
|
||||
def _om_lineage(fqn: str) -> dict[str, int]:
|
||||
"""OpenMetadata table lineage (best-effort, short timeout). Kept out of the
|
||||
hot path by default; only called when LINEAGE_OM_ENRICH=1 is set."""
|
||||
if not OPENMETADATA_URL or not fqn or os.getenv("LINEAGE_OM_ENRICH", "0") != "1":
|
||||
return {}
|
||||
url = f"{OPENMETADATA_URL}/api/v1/lineage/table/name/{fqn}?upstreamDepth=1&downstreamDepth=1"
|
||||
headers = {"Accept": "application/json"}
|
||||
if OPENMETADATA_TOKEN:
|
||||
headers["Authorization"] = f"Bearer {OPENMETADATA_TOKEN}"
|
||||
try:
|
||||
with httpx.Client(timeout=4.0, verify=False) as c:
|
||||
r = c.get(url, headers=headers)
|
||||
if r.status_code != 200:
|
||||
return {}
|
||||
j = r.json()
|
||||
return {"upstream": len(j.get("upstreamEdges") or []),
|
||||
"downstream": len(j.get("downstreamEdges") or [])}
|
||||
except Exception:
|
||||
return {}
|
||||
|
||||
|
||||
def _build() -> dict[str, Any]:
|
||||
gen_active, arch_active = _is_active()
|
||||
counts = _obs_counts()
|
||||
nodes: list[dict[str, Any]] = []
|
||||
edges: list[dict[str, Any]] = []
|
||||
seen: set[str] = set()
|
||||
|
||||
def node(nid: str, label: str, ntype: str, stage: int, **meta: Any) -> None:
|
||||
if nid in seen:
|
||||
return
|
||||
seen.add(nid)
|
||||
nodes.append({"id": nid, "label": label, "type": ntype, "stage": stage, "meta": meta})
|
||||
|
||||
def edge(src: str, dst: str, label: str, kind: str, active: bool = False) -> None:
|
||||
edges.append({"id": f"{src}->{dst}", "source": src, "target": dst,
|
||||
"label": label, "kind": kind, "active": active})
|
||||
|
||||
# Stage 5/6 shared serving nodes
|
||||
node("trino", "Trino", "engine", 5, engine="Trino", note="Federated SQL across every catalog")
|
||||
node("es", "Elasticsearch", "serving", 6, engine="Elasticsearch", note="Business + infra search indices")
|
||||
node("rag", "ChromaDB (RAG)", "serving", 6, engine="ChromaDB", note="Vector store / embeddings")
|
||||
node("vllm", "vLLM", "serving", 6, engine="vLLM", note="LLM inference (GPU)")
|
||||
node("cc", "Command Center", "serving", 6, engine="React", note="Knowledge Chat & dashboards")
|
||||
edge("rag", "vllm", "context", "serve", active=True)
|
||||
edge("vllm", "cc", "answers", "serve", active=True)
|
||||
edge("trino", "es", "index business data", "serve")
|
||||
edge("es", "cc", "search", "serve")
|
||||
edge("trino", "cc", "dashboards", "serve")
|
||||
|
||||
for key in _SOURCE_KEYS:
|
||||
ds = DATASET_BY_KEY[key]
|
||||
src_id = f"src_{key}"
|
||||
topic_id = f"topic_{key}"
|
||||
lake_id = f"lake_{key}"
|
||||
rows = _row_count(key, counts)
|
||||
cols = _pii_columns(key)
|
||||
node(src_id, f"{ds['engine']}\n{ds['table']}", "source", 0,
|
||||
engine=ds["engine"], table=ds["fqtn"], rows=rows, domain=ds.get("domain"),
|
||||
columns=cols, ts_col=ds.get("ts_col"), key_col=ds.get("key_col"),
|
||||
om=_om_lineage(ds.get("om_fqn", "")))
|
||||
|
||||
# CDC path (Debezium → Kafka topic)
|
||||
if ds.get("topic"):
|
||||
node(topic_id, f"Kafka\n{ds['topic']}", "stream", 1,
|
||||
engine="Kafka", topic=ds["topic"], note="Debezium CDC topic")
|
||||
edge(src_id, topic_id, "Debezium CDC", "cdc", active=gen_active)
|
||||
node("connect", "Kafka Connect", "stream", 2, engine="Kafka Connect",
|
||||
note="Debezium source + HDFS/S3 sinks")
|
||||
edge(topic_id, "connect", "consume", "cdc", active=gen_active)
|
||||
|
||||
# ETL offload (direct source → S3 Parquet lake)
|
||||
node(lake_id, f"S3 lake\nlake/{key}", "storage", 3, engine="ObjectScale S3",
|
||||
path=f"s3://data/lake/{key}/", note="Parquet parts (ETL offload)")
|
||||
edge(src_id, lake_id, "ETL offload (Parquet)", "batch", active=arch_active)
|
||||
|
||||
# Spark / Connect → curated + hadoop Iceberg tables
|
||||
node("spark", "Spark", "compute", 2, engine="Spark", note="Transform & mask → curated")
|
||||
edge("connect", "spark", "stream", "transform", active=gen_active)
|
||||
|
||||
cur = DATASET_BY_KEY["curated"]
|
||||
had = DATASET_BY_KEY["hadoop"]
|
||||
node("iceberg_curated", f"Iceberg\n{cur['table']}", "lakehouse", 4, engine="Iceberg",
|
||||
table=cur["fqtn"], rows=_row_count("curated", counts), masked_layer=True,
|
||||
note="PII-masked curated layer", columns=_pii_columns("curated"),
|
||||
om=_om_lineage(cur.get("om_fqn", "")))
|
||||
node("iceberg_hadoop", f"Iceberg/HDFS\n{had['table']}", "lakehouse", 4, engine="Iceberg / HDFS",
|
||||
table=had["fqtn"], rows=_row_count("hadoop", counts), note="Historical sales on HDFS",
|
||||
om=_om_lineage(had.get("om_fqn", "")))
|
||||
|
||||
edge("spark", "iceberg_curated", "mask + write", "transform", active=arch_active)
|
||||
edge("lake_orders", "iceberg_hadoop", "register external", "batch", active=arch_active)
|
||||
edge("src_orders", "iceberg_curated", "curate (masked)", "transform")
|
||||
|
||||
# Lakehouse → Trino
|
||||
for nid in ("iceberg_curated", "iceberg_hadoop"):
|
||||
edge(nid, "trino", "query", "serve")
|
||||
for key in _SOURCE_KEYS:
|
||||
edge(f"src_{key}", "trino", "federate", "serve")
|
||||
|
||||
# Lake → RAG (documents/parquet feeding embeddings is conceptual)
|
||||
edge("iceberg_curated", "rag", "embed (masked-safe)", "serve")
|
||||
|
||||
# Column-level links: source sales_orders → curated masked
|
||||
column_links: list[dict[str, Any]] = []
|
||||
src_cols = {c["name"] for c in _pii_columns("orders")}
|
||||
cur_cols = {c["name"] for c in _pii_columns("curated")}
|
||||
for cn in sorted(src_cols & cur_cols):
|
||||
masked = any(c["name"] == cn and c["masked"] for c in _pii_columns("curated"))
|
||||
column_links.append({"source": "src_orders", "target": "iceberg_curated",
|
||||
"column": cn, "masked": masked})
|
||||
|
||||
return {
|
||||
"ok": True,
|
||||
"generated_at": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()),
|
||||
"active": {"generator": gen_active, "archive": arch_active},
|
||||
"stages": ["Sources", "CDC / Stream", "Compute", "Lake storage",
|
||||
"Lakehouse", "Federation", "Serving"],
|
||||
"nodes": nodes,
|
||||
"edges": edges,
|
||||
"column_links": column_links,
|
||||
"om_connected": bool(OPENMETADATA_URL),
|
||||
}
|
||||
|
||||
|
||||
def get_graph(use_cache: bool = True) -> dict[str, Any]:
|
||||
now = time.time()
|
||||
if use_cache and _cache["data"] and now - _cache["ts"] < _TTL:
|
||||
return _cache["data"]
|
||||
data = _build()
|
||||
_cache["data"] = data
|
||||
_cache["ts"] = now
|
||||
return data
|
||||
|
||||
|
||||
@router.get("/datasets")
|
||||
async def datasets() -> JSONResponse:
|
||||
return JSONResponse({"ok": True, "datasets": [
|
||||
{"key": d["key"], "label": d["label"], "engine": d["engine"],
|
||||
"color": d["color"], "table": d["fqtn"]}
|
||||
for d in DATASETS]})
|
||||
|
||||
|
||||
@router.get("/graph")
|
||||
async def graph(dataset: str | None = None, refresh: bool = False) -> JSONResponse:
|
||||
from starlette.concurrency import run_in_threadpool
|
||||
data = await run_in_threadpool(get_graph, not refresh)
|
||||
if dataset and dataset in DATASET_BY_KEY:
|
||||
# Filter to nodes reachable from / to the selected source, keep serving spine.
|
||||
keep = {f"src_{dataset}", f"topic_{dataset}", f"lake_{dataset}",
|
||||
"connect", "spark", "iceberg_curated", "iceberg_hadoop",
|
||||
"trino", "es", "rag", "vllm", "cc"}
|
||||
nodes = [n for n in data["nodes"] if n["id"] in keep]
|
||||
node_ids = {n["id"] for n in nodes}
|
||||
edges = [e for e in data["edges"] if e["source"] in node_ids and e["target"] in node_ids]
|
||||
out = {**data, "nodes": nodes, "edges": edges, "focus": dataset}
|
||||
return JSONResponse(out)
|
||||
return JSONResponse(data)
|
||||
@@ -57,6 +57,10 @@ from spark_workbench import router as spark_workbench_router
|
||||
from pii_catalog import router as pii_router
|
||||
from trino_federated import router as federated_router
|
||||
from etl_offload import router as etl_offload_router
|
||||
from lineage import router as lineage_router
|
||||
from dq_monitor import router as dq_router
|
||||
from observability import router as observability_router
|
||||
from catalog_governance import router as governance_router
|
||||
from ssh_terminal import ssh_session
|
||||
from node_registry import NODE_IDS, NODE_AGENT, NODE_REGISTRY, is_node_id
|
||||
from node_ops import build_node_detail, probe_node, run_node_probe_task
|
||||
@@ -768,6 +772,10 @@ app.include_router(spark_workbench_router)
|
||||
app.include_router(pii_router)
|
||||
app.include_router(federated_router)
|
||||
app.include_router(etl_offload_router)
|
||||
app.include_router(lineage_router)
|
||||
app.include_router(dq_router)
|
||||
app.include_router(observability_router)
|
||||
app.include_router(governance_router)
|
||||
app.add_middleware(
|
||||
CORSMiddleware,
|
||||
allow_origins=["*"],
|
||||
|
||||
@@ -0,0 +1,370 @@
|
||||
"""Data observability (Disease #6 — Traceability / observability).
|
||||
|
||||
A lightweight background monitor that, per business table, tracks over time:
|
||||
|
||||
volume – row count and its delta between cycles
|
||||
freshness – age of the newest record
|
||||
schema – the column set; a change raises a drift alert
|
||||
|
||||
It derives alerts (critical/warning/info) for volume drops, stalled ingestion,
|
||||
stale data and schema drift, and keeps a rolling time-series per dataset for the
|
||||
trend charts in the UI.
|
||||
|
||||
Endpoints:
|
||||
GET /api/observability/metrics -> per-dataset series + current state
|
||||
GET /api/observability/alerts -> active + recent alerts
|
||||
POST /api/observability/run -> run one probe cycle now
|
||||
POST /api/observability/config -> {enabled, interval_s, freshness_min}
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import threading
|
||||
import time
|
||||
import uuid
|
||||
from collections import deque
|
||||
from datetime import datetime, timezone
|
||||
from typing import Any
|
||||
|
||||
from fastapi import APIRouter, Body
|
||||
from fastapi.responses import JSONResponse
|
||||
|
||||
from lake_meta import DATASETS, trino_scalar, discover_columns
|
||||
|
||||
router = APIRouter(prefix="/api/observability", tags=["observability"])
|
||||
|
||||
_lock = threading.Lock()
|
||||
_state: dict[str, Any] = {
|
||||
"enabled": True,
|
||||
"interval_s": 90.0,
|
||||
"freshness_min": 30.0,
|
||||
"running": False,
|
||||
"cycles": 0,
|
||||
"last_cycle_ts": 0.0,
|
||||
"next_run_ts": 0.0,
|
||||
"started_at": None,
|
||||
"datasets": {
|
||||
d["key"]: {"label": d["label"], "engine": d["engine"], "color": d["color"],
|
||||
"table": d["fqtn"], "rows": None, "prev_rows": None, "delta": None,
|
||||
"freshness_age_min": None, "columns": None, "schema_hash": None,
|
||||
"series": deque(maxlen=80), "deltas": deque(maxlen=20),
|
||||
"stalled_cycles": 0, "ts": None, "error": None}
|
||||
for d in DATASETS
|
||||
},
|
||||
"alerts_active": {}, # dedup_key -> alert
|
||||
"alerts_history": deque(maxlen=120),
|
||||
"feed": deque(maxlen=40),
|
||||
}
|
||||
|
||||
|
||||
def _now() -> datetime:
|
||||
return datetime.now(timezone.utc)
|
||||
|
||||
|
||||
def _term(text: str, level: str = "info", phase: str = "observe") -> None:
|
||||
try:
|
||||
from agent_terminal import emit_threadsafe
|
||||
emit_threadsafe("infra-sentinel", text, level=level, phase=phase)
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
|
||||
def _raise_alert(dataset: str, atype: str, severity: str, message: str) -> None:
|
||||
dedup = f"{dataset}:{atype}"
|
||||
existing = _state["alerts_active"].get(dedup)
|
||||
if existing:
|
||||
existing["count"] += 1
|
||||
existing["last_ts"] = _now().isoformat()
|
||||
existing["message"] = message
|
||||
return
|
||||
alert = {"id": uuid.uuid4().hex[:8], "dataset": dataset, "type": atype,
|
||||
"severity": severity, "message": message, "count": 1,
|
||||
"ts": _now().isoformat(), "last_ts": _now().isoformat(), "resolved": False}
|
||||
_state["alerts_active"][dedup] = alert
|
||||
_state["alerts_history"].appendleft(dict(alert))
|
||||
lvl = "err" if severity == "critical" else ("warn" if severity == "warning" else "info")
|
||||
_term(f" ⚠ ALERT [{severity}] {dataset}: {message}", level=lvl, phase="alert")
|
||||
_state["feed"].appendleft({"ts": _now().isoformat(), "text": f"[{severity}] {dataset}: {message}", "level": lvl})
|
||||
|
||||
|
||||
def _clear_alert(dataset: str, atype: str) -> None:
|
||||
dedup = f"{dataset}:{atype}"
|
||||
a = _state["alerts_active"].pop(dedup, None)
|
||||
if a:
|
||||
a["resolved"] = True
|
||||
a["resolved_ts"] = _now().isoformat()
|
||||
_state["alerts_history"].appendleft({**a, "message": f"resolved: {a['message']}"})
|
||||
|
||||
|
||||
def _native_freshness(ds: dict[str, Any]) -> float | None:
|
||||
"""Epoch seconds of the newest record, read cheaply via the PK index
|
||||
(ORDER BY <pk> DESC LIMIT 1) on the source DB. Returns None if unavailable."""
|
||||
nc = ds.get("native_count")
|
||||
ts = ds.get("ts_col")
|
||||
key = ds.get("key_col")
|
||||
if not nc or not ts:
|
||||
return None
|
||||
eng = nc[0]
|
||||
try:
|
||||
import sql_console as s
|
||||
if eng == "postgres" and key:
|
||||
import psycopg2
|
||||
conn = psycopg2.connect(host=s.DB_HOST, port=s.PG_PORT, user=s.PG_USER,
|
||||
password=s.PG_PASS, dbname=s.PG_DB, connect_timeout=6)
|
||||
try:
|
||||
cur = conn.cursor()
|
||||
cur.execute("SET statement_timeout = 6000")
|
||||
cur.execute(f'SELECT extract(epoch FROM "{ts}") FROM public."{ds["table"]}" '
|
||||
f'WHERE "{ts}" IS NOT NULL ORDER BY "{key}" DESC LIMIT 1')
|
||||
row = cur.fetchone()
|
||||
return float(row[0]) if row and row[0] is not None else None
|
||||
finally:
|
||||
conn.close()
|
||||
if eng == "mysql" and key:
|
||||
import pymysql
|
||||
conn = pymysql.connect(host=s.DB_HOST, port=s.MYSQL_PORT, user=s.MYSQL_USER,
|
||||
password=s.MYSQL_PASS, database=s.MYSQL_DB, connect_timeout=6)
|
||||
try:
|
||||
cur = conn.cursor()
|
||||
cur.execute("SET SESSION MAX_EXECUTION_TIME = 6000")
|
||||
cur.execute(f"SELECT UNIX_TIMESTAMP(`{ts}`) FROM `{ds['table']}` "
|
||||
f"WHERE `{ts}` IS NOT NULL ORDER BY `{key}` DESC LIMIT 1")
|
||||
row = cur.fetchone()
|
||||
return float(row[0]) if row and row[0] is not None else None
|
||||
finally:
|
||||
conn.close()
|
||||
if eng == "mongodb":
|
||||
cli = s._mongo_client()
|
||||
try:
|
||||
doc = list(cli[ds["schema"]][ds["table"]].find({}, {ts: 1, "_id": 0}).sort("_id", -1).limit(1))
|
||||
if doc:
|
||||
import datetime as _dt
|
||||
v = doc[0].get(ts)
|
||||
if isinstance(v, _dt.datetime):
|
||||
return v.timestamp()
|
||||
finally:
|
||||
cli.close()
|
||||
except Exception:
|
||||
return None
|
||||
return None
|
||||
|
||||
|
||||
def _probe(ds: dict[str, Any]) -> None:
|
||||
key = ds["key"]
|
||||
st = _state["datasets"][key]
|
||||
fq = ds["fqtn"]
|
||||
|
||||
# volume — prefer a fast native planner estimate (postgres/mysql/mongo); use
|
||||
# Trino metadata count for Iceberg; treat a slow/failed count as a soft miss.
|
||||
rows = None
|
||||
estimated = False
|
||||
nc = ds.get("native_count")
|
||||
if nc:
|
||||
try:
|
||||
import sql_console as _s
|
||||
rows = _s._table_row_count(nc[0], nc[1])
|
||||
estimated = True
|
||||
except Exception:
|
||||
rows = None
|
||||
if rows is None:
|
||||
try:
|
||||
rows = int(trino_scalar(f"SELECT count(*) FROM {fq}", timeout=30.0) or 0)
|
||||
except Exception as exc:
|
||||
st["error"] = str(exc)[:140]
|
||||
st["count_fails"] = st.get("count_fails", 0) + 1
|
||||
return
|
||||
st["error"] = None
|
||||
st["count_fails"] = 0
|
||||
st["estimated"] = estimated
|
||||
|
||||
prev = st["rows"]
|
||||
st["prev_rows"] = prev
|
||||
st["rows"] = rows
|
||||
delta = (rows - prev) if prev is not None else None
|
||||
st["delta"] = delta
|
||||
st["series"].append({"t": _now().strftime("%H:%M:%S"), "rows": rows, "delta": delta or 0})
|
||||
st["ts"] = _now().isoformat()
|
||||
|
||||
if delta is not None:
|
||||
st["deltas"].append(delta)
|
||||
# Guard against planner-estimate jitter: only alert on a material drop.
|
||||
drop_threshold = max(2000, int(0.02 * (prev or 0)))
|
||||
if delta < -drop_threshold:
|
||||
_raise_alert(key, "volume_drop", "warning",
|
||||
f"row count dropped by {abs(delta):,} ({prev:,} → {rows:,})")
|
||||
else:
|
||||
_clear_alert(key, "volume_drop")
|
||||
# stalled ingestion: no growth for several cycles on CDC sources. Skip for
|
||||
# estimate-based counts (planner stats update lazily → false stalls).
|
||||
if delta == 0 and not ds.get("curated") and not estimated:
|
||||
st["stalled_cycles"] += 1
|
||||
if st["stalled_cycles"] >= 4:
|
||||
_raise_alert(key, "stalled", "warning",
|
||||
f"no new rows for {st['stalled_cycles']} cycles")
|
||||
else:
|
||||
st["stalled_cycles"] = 0
|
||||
_clear_alert(key, "stalled")
|
||||
# spike (informational) — only for exact counts
|
||||
pos = [d for d in st["deltas"] if d > 0]
|
||||
if not estimated and pos and delta > (sum(pos) / len(pos)) * 6 and len(pos) >= 4:
|
||||
_raise_alert(key, "spike", "info", f"volume spike +{delta:,} rows this cycle")
|
||||
|
||||
# freshness — cheap PK-indexed newest-row lookup for the CDC sources, with a
|
||||
# short Trino max(ts) fallback for Iceberg tables.
|
||||
if ds.get("ts_col"):
|
||||
epoch = _native_freshness(ds)
|
||||
if epoch is None:
|
||||
try:
|
||||
v = trino_scalar(f'SELECT to_unixtime(max("{ds["ts_col"]}")) FROM {fq}', timeout=10.0)
|
||||
epoch = float(v) if v is not None else None
|
||||
except Exception:
|
||||
epoch = None
|
||||
if epoch:
|
||||
age = max(0.0, (time.time() - float(epoch)) / 60.0)
|
||||
st["freshness_age_min"] = round(age, 1)
|
||||
if age > float(_state["freshness_min"]) and not ds.get("curated"):
|
||||
_raise_alert(key, "stale", "warning",
|
||||
f"newest record is {age:.0f}m old (> {_state['freshness_min']:.0f}m)")
|
||||
else:
|
||||
_clear_alert(key, "stale")
|
||||
|
||||
# schema drift
|
||||
cols = discover_columns(ds)
|
||||
if cols:
|
||||
names = sorted(c["name"] for c in cols)
|
||||
h = hash(tuple(names))
|
||||
if st["schema_hash"] is not None and h != st["schema_hash"]:
|
||||
old = set(st["columns"] or [])
|
||||
new = set(names)
|
||||
added = sorted(new - old)
|
||||
removed = sorted(old - new)
|
||||
parts = []
|
||||
if added:
|
||||
parts.append(f"+{', '.join(added)}")
|
||||
if removed:
|
||||
parts.append(f"-{', '.join(removed)}")
|
||||
_raise_alert(key, "schema_drift", "critical", "schema changed: " + " ".join(parts))
|
||||
st["columns"] = names
|
||||
st["schema_hash"] = h
|
||||
|
||||
|
||||
def run_cycle() -> dict[str, Any]:
|
||||
if _state["running"]:
|
||||
return {"ok": True, "skipped": "already running"}
|
||||
_state["running"] = True
|
||||
try:
|
||||
_term(f"═══ observability sweep {_state['cycles'] + 1} — volume · freshness · schema ═══",
|
||||
level="info", phase="cycle")
|
||||
for ds in DATASETS:
|
||||
if not _state["enabled"]:
|
||||
break
|
||||
try:
|
||||
_probe(ds)
|
||||
except Exception as exc:
|
||||
_state["datasets"][ds["key"]]["error"] = str(exc)[:140]
|
||||
_state["cycles"] += 1
|
||||
_state["last_cycle_ts"] = time.time()
|
||||
_state["next_run_ts"] = time.time() + float(_state["interval_s"])
|
||||
active = len(_state["alerts_active"])
|
||||
_term(f"═══ sweep {_state['cycles']} done — {active} active alert(s) ═══",
|
||||
level=("warn" if active else "ok"), phase="cycle")
|
||||
finally:
|
||||
_state["running"] = False
|
||||
return {"ok": True, "active_alerts": len(_state["alerts_active"])}
|
||||
|
||||
|
||||
def _loop() -> None:
|
||||
_state["started_at"] = _now().isoformat()
|
||||
time.sleep(28)
|
||||
while True:
|
||||
try:
|
||||
if _state["enabled"]:
|
||||
run_cycle()
|
||||
except Exception:
|
||||
pass
|
||||
time.sleep(max(30.0, float(_state["interval_s"])))
|
||||
|
||||
|
||||
threading.Thread(target=_loop, daemon=True, name="observability").start()
|
||||
|
||||
|
||||
def metrics_view() -> dict[str, Any]:
|
||||
with _lock:
|
||||
datasets = []
|
||||
for ds in DATASETS:
|
||||
st = _state["datasets"][ds["key"]]
|
||||
datasets.append({
|
||||
"key": ds["key"], "label": st["label"], "engine": st["engine"],
|
||||
"color": st["color"], "table": st["table"], "rows": st["rows"],
|
||||
"delta": st["delta"], "freshness_age_min": st["freshness_age_min"],
|
||||
"columns": len(st["columns"]) if st["columns"] else None,
|
||||
"stalled_cycles": st["stalled_cycles"], "error": st["error"],
|
||||
"ts": st["ts"], "series": list(st["series"]),
|
||||
})
|
||||
return {
|
||||
"ok": True, "enabled": _state["enabled"], "interval_s": _state["interval_s"],
|
||||
"freshness_min": _state["freshness_min"], "running": _state["running"],
|
||||
"cycles": _state["cycles"], "last_cycle_ts": _state["last_cycle_ts"],
|
||||
"next_run_ts": _state["next_run_ts"],
|
||||
"alert_counts": _alert_counts(),
|
||||
"datasets": datasets, "feed": list(_state["feed"])[:20],
|
||||
}
|
||||
|
||||
|
||||
def _alert_counts() -> dict[str, int]:
|
||||
counts = {"critical": 0, "warning": 0, "info": 0}
|
||||
for a in _state["alerts_active"].values():
|
||||
counts[a["severity"]] = counts.get(a["severity"], 0) + 1
|
||||
counts["total"] = len(_state["alerts_active"])
|
||||
return counts
|
||||
|
||||
|
||||
def summary_for_llm() -> dict[str, Any]:
|
||||
with _lock:
|
||||
return {
|
||||
"active_alerts": _alert_counts(),
|
||||
"alerts": [{"dataset": a["dataset"], "type": a["type"], "severity": a["severity"],
|
||||
"message": a["message"]} for a in _state["alerts_active"].values()],
|
||||
}
|
||||
|
||||
|
||||
@router.get("/metrics")
|
||||
async def get_metrics() -> JSONResponse:
|
||||
return JSONResponse(metrics_view())
|
||||
|
||||
|
||||
@router.get("/alerts")
|
||||
async def get_alerts() -> JSONResponse:
|
||||
with _lock:
|
||||
return JSONResponse({
|
||||
"ok": True,
|
||||
"counts": _alert_counts(),
|
||||
"active": sorted(_state["alerts_active"].values(),
|
||||
key=lambda a: {"critical": 0, "warning": 1, "info": 2}.get(a["severity"], 3)),
|
||||
"history": list(_state["alerts_history"])[:60],
|
||||
})
|
||||
|
||||
|
||||
@router.post("/run")
|
||||
async def post_run() -> JSONResponse:
|
||||
from starlette.concurrency import run_in_threadpool
|
||||
res = await run_in_threadpool(run_cycle)
|
||||
return JSONResponse({**res, "metrics": metrics_view()})
|
||||
|
||||
|
||||
@router.post("/config")
|
||||
async def post_config(body: dict = Body(default={})) -> JSONResponse:
|
||||
if "enabled" in body:
|
||||
_state["enabled"] = bool(body["enabled"])
|
||||
if "interval_s" in body:
|
||||
try:
|
||||
_state["interval_s"] = max(30.0, min(900.0, float(body["interval_s"])))
|
||||
except Exception:
|
||||
pass
|
||||
if "freshness_min" in body:
|
||||
try:
|
||||
_state["freshness_min"] = max(1.0, min(1440.0, float(body["freshness_min"])))
|
||||
except Exception:
|
||||
pass
|
||||
return JSONResponse({"ok": True, "enabled": _state["enabled"],
|
||||
"interval_s": _state["interval_s"], "freshness_min": _state["freshness_min"]})
|
||||
Reference in New Issue
Block a user