stackpilot/backend/services/auto_update_service.py

"""Auto-update (Watchtower-style) orchestration.

Runs once per image-update-check cycle (called from
``update_service.background_loop`` so it reuses the freshly-populated digest
cache). For each enabled policy whose stack has a newer image available, either
pulls + redeploys the stack or just notifies, recording the outcome.

Central-only / DB-aware. Image resolution + digest comparison live in the
DB-free ``update_service`` so the agent can answer ``/agent/stacks/{id}/updates``
with the same logic.
"""
from __future__ import annotations

import logging
from datetime import datetime, timezone

from sqlmodel import Session, select

from database import engine
from models.agent import Agent
from models.auto_update import AutoUpdate
from models.setting import EVENT_PULL_FAILED, EVENT_STACK_AUTO_UPDATED
from services import (
    agent_service,
    compose_service,
    notify_service,
    stack_lock_service,
    update_service,
)

logger = logging.getLogger("stackpilot.autoupdate")

# Stack states we'll act on. We never auto-start a stopped stack.
_LIVE_STATES = {"running", "partial", "updating"}


def _now() -> datetime:
    return datetime.now(timezone.utc)


# --------------------------------------------------------------------------- #
# Policy CRUD helpers (shared by the stacks + agents routers)
# --------------------------------------------------------------------------- #


def get_policy(session: Session, stack_id: str, agent_id: int | None = None) -> AutoUpdate | None:
    stmt = select(AutoUpdate).where(AutoUpdate.stack_id == stack_id)
    stmt = stmt.where(AutoUpdate.agent_id == agent_id) if agent_id is not None \
        else stmt.where(AutoUpdate.agent_id.is_(None))
    return session.exec(stmt).first()


def upsert_policy(
    session: Session, stack_id: str, enabled: bool, redeploy: bool, agent_id: int | None = None
) -> AutoUpdate:
    policy = get_policy(session, stack_id, agent_id)
    if policy is None:
        policy = AutoUpdate(stack_id=stack_id, agent_id=agent_id)
    policy.enabled = enabled
    policy.redeploy = redeploy
    session.add(policy)
    session.commit()
    session.refresh(policy)
    return policy


def to_read(session: Session, policy: AutoUpdate | None, stack_id: str, agent_id: int | None = None) -> dict:
    """Build an AutoUpdateRead-shaped dict, defaulting to disabled when absent."""
    agent_name = None
    if agent_id is not None:
        agent = session.get(Agent, agent_id)
        agent_name = agent.name if agent else None
    if policy is None:
        return {
            "id": None, "stack_id": stack_id, "agent_id": agent_id, "agent_name": agent_name,
            "enabled": False, "redeploy": True,
            "last_run": None, "last_status": None, "last_result": None,
        }
    return {
        "id": policy.id, "stack_id": policy.stack_id, "agent_id": policy.agent_id,
        "agent_name": agent_name, "enabled": policy.enabled, "redeploy": policy.redeploy,
        "last_run": policy.last_run, "last_status": policy.last_status, "last_result": policy.last_result,
    }


def _record(session: Session, policy: AutoUpdate, status: str, result: str = "") -> None:
    policy.last_run = _now()
    policy.last_status = status
    policy.last_result = result[:300] if result else None
    session.add(policy)
    session.commit()


async def _run_local(session: Session, policy: AutoUpdate) -> None:
    stack_id = policy.stack_id
    try:
        state = compose_service.compute_status(stack_id)
    except Exception:  # noqa: BLE001
        state = "unknown"
    if state not in _LIVE_STATES:
        _record(session, policy, "skipped", f"stack not running ({state})")
        return

    summary = await update_service.stack_updates(stack_id, refresh=False)
    if not summary["update_available"]:
        _record(session, policy, "up-to-date")
        return

    stale = ", ".join(summary["stale_images"])
    prev = policy.last_status
    if policy.redeploy:
        try:
            # Never redeploy underneath somebody: if a user is mid-deploy on
            # this stack, skip and pick it up next cycle rather than racing
            # them over the same containers.
            with stack_lock_service.hold(session, stack_id, "auto-update", "auto-update"):
                await compose_service.pull(stack_id)
                await compose_service.up(stack_id)
        except stack_lock_service.StackBusy as exc:
            _record(session, policy, "skipped", str(exc))
            return
        except Exception as exc:  # noqa: BLE001
            _record(session, policy, "error", str(exc))
            await _safe_notify(EVENT_PULL_FAILED, f"Auto-update of '{stack_id}' failed", str(exc), session)
            return
        update_service.refresh_stack_local(stack_id)
        _record(session, policy, "updated", stale)
        await _safe_notify(
            EVENT_STACK_AUTO_UPDATED, f"Stack '{stack_id}' auto-updated",
            f"Pulled and redeployed: {stale}.", session,
        )
    else:
        _record(session, policy, "update-available", stale)
        if prev != "update-available":  # notify once per transition, not every cycle
            await _safe_notify(
                EVENT_STACK_AUTO_UPDATED, f"Update available for '{stack_id}'",
                f"Newer images: {stale} (auto-redeploy is off).", session,
            )


async def _run_remote(session: Session, policy: AutoUpdate) -> None:
    agent = session.get(Agent, policy.agent_id)
    if not agent:
        _record(session, policy, "error", "agent not found")
        return
    stack_id = policy.stack_id
    try:
        summary = await agent_service.call(
            session, agent, "GET", f"/agent/stacks/{stack_id}/updates",
            params={"refresh": "true"},
        )
    except Exception as exc:  # noqa: BLE001
        _record(session, policy, "error", f"agent check failed: {exc}")
        return
    if not summary or not summary.get("update_available"):
        _record(session, policy, "up-to-date")
        return

    stale = ", ".join(summary.get("stale_images", []))
    label = f"{agent.name}/{stack_id}"
    prev = policy.last_status
    if policy.redeploy:
        try:
            await agent_service.call(session, agent, "POST", f"/agent/stacks/{stack_id}/update")
        except Exception as exc:  # noqa: BLE001
            _record(session, policy, "error", str(exc))
            await _safe_notify(EVENT_PULL_FAILED, f"Auto-update of '{label}' failed", str(exc), session)
            return
        _record(session, policy, "updated", stale)
        await _safe_notify(
            EVENT_STACK_AUTO_UPDATED, f"Stack '{label}' auto-updated",
            f"Pulled and redeployed: {stale}.", session,
        )
    else:
        _record(session, policy, "update-available", stale)
        if prev != "update-available":
            await _safe_notify(
                EVENT_STACK_AUTO_UPDATED, f"Update available for '{label}'",
                f"Newer images: {stale} (auto-redeploy is off).", session,
            )


async def _safe_notify(event: str, title: str, message: str, session: Session) -> None:
    try:
        await notify_service.notify(event, title, message, session)
    except Exception as exc:  # noqa: BLE001 - notifications are best-effort
        logger.debug("auto-update notify failed: %s", exc)


async def run_policy(session: Session, policy: AutoUpdate) -> None:
    if policy.agent_id is None:
        await _run_local(session, policy)
    else:
        await _run_remote(session, policy)


async def run_due() -> None:
    """Process every enabled policy. Best-effort: one failure never aborts the rest."""
    with Session(engine) as session:
        policies = list(session.exec(select(AutoUpdate).where(AutoUpdate.enabled == True)))  # noqa: E712
    for policy in policies:
        try:
            with Session(engine) as session:
                fresh = session.get(AutoUpdate, policy.id)
                if fresh and fresh.enabled:
                    await run_policy(session, fresh)
        except Exception as exc:  # noqa: BLE001
            logger.warning("Auto-update policy %s failed: %s", policy.id, exc)