guardia-itsm/core/deploy_pipeline.py

"""
core/deploy_pipeline.py — 배포 파이프라인 오케스트레이터.

빌드 → 테스트 → 배포 → 헬스체크 → ITSM 콜백의 전체 파이프라인을 관리한다.

단계별 진행 상황은 tb_vibe_session.status 와 work_log에 기록된다.
실패 시 자동 롤백을 수행하고 SR 담당자에게 메신저 알림을 발송한다.

보안:
  - SSH 실행은 core/ssh_exec.py의 execute_ssh_command() 경유 (명령어 안전성 검증 포함)
  - 서버 자격증명(IP, SSH 계정) API 응답에 포함 금지
  - 헬스체크 URL은 내부 네트워크 도메인만 허용
"""
from __future__ import annotations

import asyncio
import logging
import os
from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from typing import Any, Optional

import httpx

logger = logging.getLogger(__name__)

# ── 환경변수 ──────────────────────────────────────────────────────────────────
_HEALTH_CHECK_RETRIES = int(os.getenv("DEPLOY_HEALTH_RETRIES", "10"))
_HEALTH_CHECK_INTERVAL = float(os.getenv("DEPLOY_HEALTH_INTERVAL", "10"))  # 초
_DEPLOY_TIMEOUT = int(os.getenv("DEPLOY_TIMEOUT", "1800"))                 # 30분


# ── 파이프라인 단계 ───────────────────────────────────────────────────────────

class PipelineStage(str, Enum):
    PRE_CHECK    = "PRE_CHECK"
    BUILD        = "BUILD"
    TEST         = "TEST"
    BACKUP       = "BACKUP"
    DEPLOY       = "DEPLOY"
    RESTART      = "RESTART"
    HEALTH_CHECK = "HEALTH_CHECK"
    NOTIFY       = "NOTIFY"
    COMPLETED    = "COMPLETED"
    FAILED       = "FAILED"
    ROLLED_BACK  = "ROLLED_BACK"


@dataclass
class StageResult:
    stage:    PipelineStage
    success:  bool
    message:  str
    stdout:   str = ""
    duration: float = 0.0     # 초


@dataclass
class PipelineResult:
    session_id:  int
    success:     bool
    final_stage: PipelineStage
    stages:      list[StageResult] = field(default_factory=list)
    rollback_done: bool = False
    error_msg:   str = ""

    @property
    def summary(self) -> str:
        lines = [f"파이프라인 {'성공' if self.success else '실패'} — 세션 {self.session_id}"]
        for r in self.stages:
            icon = "✅" if r.success else "❌"
            lines.append(f"  {icon} {r.stage.value}: {r.message} ({r.duration:.1f}s)")
        if self.rollback_done:
            lines.append("  ⏪ 자동 롤백 완료")
        return "\n".join(lines)


@dataclass
class PipelineStatus:
    session_id:  int
    current_stage: PipelineStage
    completed_stages: list[str]
    started_at:  Optional[datetime]
    is_running:  bool


# ── DeployPipeline ────────────────────────────────────────────────────────────

class DeployPipeline:
    """
    8단계 배포 파이프라인 오케스트레이터.

    사용 예시:
        pipeline = DeployPipeline()
        result   = await pipeline.run(session_id=42)
        if not result.success:
            await pipeline.rollback(session_id=42)
    """

    def __init__(self) -> None:
        self._running: dict[int, PipelineStatus] = {}

    # ── 파이프라인 실행 ────────────────────────────────────────────────────────

    async def run(self, session_id: int) -> PipelineResult:
        """
        전체 배포 파이프라인을 실행한다.

        단계별 실패 처리:
          - test 단계 실패  → 배포 중단, SR 상태 FAILED_TEST
          - health_check 10회 실패 → 자동 rollback 실행
          - 기타 단계 실패 → 배포 중단 후 에러 보고

        Args:
            session_id: tb_vibe_session.id

        Returns:
            PipelineResult
        """
        from database import SessionLocal
        from models import VibeSession, VibeSessionStatus, WorkLog

        result = PipelineResult(session_id=session_id, success=False, final_stage=PipelineStage.PRE_CHECK)
        stages: list[StageResult] = []

        self._running[session_id] = PipelineStatus(
            session_id=session_id,
            current_stage=PipelineStage.PRE_CHECK,
            completed_stages=[],
            started_at=datetime.utcnow(),
            is_running=True,
        )

        async with SessionLocal() as db:
            session = await db.get(VibeSession, session_id)
            if not session:
                result.error_msg = f"세션을 찾을 수 없습니다: {session_id}"
                return result

            project = session.project

        try:
            # 1. 사전 점검
            sr = await self._pre_check(session_id, project)
            stages.append(sr)
            self._advance(session_id, PipelineStage.BUILD, sr)
            if not sr.success:
                result.error_msg = sr.message
                result.final_stage = PipelineStage.FAILED
                await self._update_session(session_id, "FAILED", sr.message)
                return result

            # 2. 빌드
            sr = await self._build(session_id, project)
            stages.append(sr)
            self._advance(session_id, PipelineStage.TEST, sr)
            if not sr.success:
                result.error_msg = sr.message
                result.final_stage = PipelineStage.FAILED
                await self._update_session(session_id, "FAILED", sr.message)
                return result

            # 빌드 완료 기록
            await self._update_session(session_id, "BUILDING", sr.message, mark_built=True)

            # 3. 테스트
            sr = await self._test(session_id, project)
            stages.append(sr)
            if not sr.success:
                result.error_msg = sr.message
                result.final_stage = PipelineStage.FAILED
                await self._update_session(session_id, "FAILED_TEST", sr.message)
                await self._notify(session_id, False, f"테스트 실패로 배포 중단\n{sr.message}")
                return result

            await self._update_session(session_id, "TESTING", sr.message, mark_tested=True)
            self._advance(session_id, PipelineStage.BACKUP, sr)

            # 4. 백업
            sr = await self._backup(session_id, project)
            stages.append(sr)
            self._advance(session_id, PipelineStage.DEPLOY, sr)
            # 백업 실패는 경고만 (중단하지 않음)
            if not sr.success:
                logger.warning("배포 백업 실패 (계속 진행): session_id=%d msg=%s", session_id, sr.message)

            # 5. 배포
            sr = await self._deploy(session_id, project)
            stages.append(sr)
            self._advance(session_id, PipelineStage.RESTART, sr)
            if not sr.success:
                result.error_msg = sr.message
                result.final_stage = PipelineStage.FAILED
                await self._update_session(session_id, "DEPLOYING", sr.message)
                return result

            await self._update_session(session_id, "DEPLOYING", sr.message)

            # 6. 서비스 재시작
            sr = await self._restart(session_id, project)
            stages.append(sr)
            self._advance(session_id, PipelineStage.HEALTH_CHECK, sr)
            if not sr.success:
                logger.warning("서비스 재시작 실패 (헬스체크 진행): %s", sr.message)

            # 7. 헬스체크
            sr = await self._health_check(session_id, project)
            stages.append(sr)
            if not sr.success:
                # 자동 롤백
                logger.error("헬스체크 실패 — 자동 롤백 시작: session_id=%d", session_id)
                rolled_back = await self.rollback(session_id)
                result.rollback_done = rolled_back
                result.error_msg = f"헬스체크 실패 (자동 롤백: {'성공' if rolled_back else '실패'})"
                result.final_stage = PipelineStage.ROLLED_BACK if rolled_back else PipelineStage.FAILED
                await self._update_session(session_id, "FAILED_ROLLBACK", result.error_msg)
                await self._notify(session_id, False, result.error_msg)
                return result

            # 8. 알림
            sr = await self._notify(session_id, True, "배포 성공")
            stages.append(sr)

            # 완료
            result.success = True
            result.final_stage = PipelineStage.COMPLETED
            await self._update_session(session_id, "COMPLETED", "배포 완료", mark_deployed=True)
            logger.info("배포 파이프라인 완료: session_id=%d", session_id)

        except Exception as exc:
            result.error_msg = str(exc)[:500]
            result.final_stage = PipelineStage.FAILED
            await self._update_session(session_id, "FAILED", result.error_msg)
            logger.exception("배포 파이프라인 예외: session_id=%d", session_id)

        finally:
            result.stages = stages
            status = self._running.pop(session_id, None)
            if status:
                status.is_running = False

        return result

    # ── 롤백 ────────────────────────────────────────────────────────────────

    async def rollback(self, session_id: int) -> bool:
        """
        이전 배포본으로 롤백한다.

        백업 디렉터리에서 파일을 복원하고 서비스를 재시작한다.

        Returns:
            True: 롤백 성공, False: 실패
        """
        from database import SessionLocal
        from models import VibeSession

        logger.info("배포 롤백 시작: session_id=%d", session_id)
        async with SessionLocal() as db:
            session = await db.get(VibeSession, session_id)
            if not session or not session.project:
                return False

            project = session.project

        try:
            project_name = getattr(project, "project_name", None) or getattr(project, "name", "")
            deploy_path = getattr(project, "deploy_path", "")
            restart_cmd = getattr(project, "was_restart_cmd", "")

            if not deploy_path:
                logger.error("롤백 실패: deploy_path 미설정 session_id=%d", session_id)
                return False

            # 백업본 복원
            backup_path = f"{deploy_path}.backup"
            from core.ssh_exec import execute_ssh_command
            server_id = getattr(project, "server_id", None)
            if server_id:
                restore_cmd = f"cp -rf {backup_path}/* {deploy_path}/ 2>/dev/null && echo RESTORED"
                result = await execute_ssh_command(server_id, restore_cmd, timeout=300)
                if result.exit_code != 0:
                    logger.error("백업 복원 실패: %s", result.stdout[:200])
                    return False

                if restart_cmd:
                    await execute_ssh_command(server_id, restart_cmd, timeout=120)

            logger.info("롤백 완료: session_id=%d", session_id)
            return True

        except Exception as exc:
            logger.exception("롤백 예외: session_id=%d err=%s", session_id, exc)
            return False

    # ── 상태 조회 ────────────────────────────────────────────────────────────

    async def get_status(self, session_id: int) -> PipelineStatus:
        """현재 파이프라인 진행 단계를 반환한다."""
        status = self._running.get(session_id)
        if status:
            return status
        # 실행 중 아닌 경우 DB에서 조회
        from database import SessionLocal
        from models import VibeSession
        async with SessionLocal() as db:
            session = await db.get(VibeSession, session_id)
            stage_str = (session.status if session else "COMPLETED") or "COMPLETED"
        return PipelineStatus(
            session_id=session_id,
            current_stage=PipelineStage(stage_str) if stage_str in PipelineStage.__members__ else PipelineStage.COMPLETED,
            completed_stages=[],
            started_at=None,
            is_running=False,
        )

    # ── 단계별 구현 ──────────────────────────────────────────────────────────

    async def _pre_check(self, session_id: int, project: Any) -> StageResult:
        """배포 환경 사전 점검: 서버 연결, 디스크 용량."""
        start = datetime.utcnow()
        try:
            from core.ssh_exec import execute_ssh_command
            server_id = getattr(project, "server_id", None)
            if not server_id:
                return StageResult(PipelineStage.PRE_CHECK, False, "server_id 미설정", duration=_elapsed(start))

            result = await execute_ssh_command(
                server_id,
                "df -h / | tail -1 | awk '{print $5}' | sed 's/%//'",
                timeout=30,
            )
            disk_pct = int(result.stdout.strip() or "0")
            if disk_pct >= 90:
                return StageResult(
                    PipelineStage.PRE_CHECK, False,
                    f"디스크 용량 부족: {disk_pct}% 사용 중",
                    duration=_elapsed(start),
                )
            return StageResult(PipelineStage.PRE_CHECK, True, f"사전 점검 통과 (디스크 {disk_pct}%)", duration=_elapsed(start))
        except Exception as exc:
            return StageResult(PipelineStage.PRE_CHECK, False, f"사전 점검 오류: {exc}", duration=_elapsed(start))

    async def _build(self, session_id: int, project: Any) -> StageResult:
        """build_cmd 실행."""
        start = datetime.utcnow()
        build_cmd = getattr(project, "build_cmd", None)
        if not build_cmd:
            return StageResult(PipelineStage.BUILD, True, "빌드 명령어 미설정 (생략)", duration=_elapsed(start))
        try:
            from core.ssh_exec import execute_ssh_command
            server_id = getattr(project, "server_id", None)
            result = await execute_ssh_command(server_id, build_cmd, timeout=_DEPLOY_TIMEOUT)
            ok = result.exit_code == 0
            return StageResult(
                PipelineStage.BUILD, ok,
                "빌드 성공" if ok else f"빌드 실패 (rc={result.exit_code})",
                stdout=result.stdout[-3000:],
                duration=_elapsed(start),
            )
        except Exception as exc:
            return StageResult(PipelineStage.BUILD, False, str(exc), duration=_elapsed(start))

    async def _test(self, session_id: int, project: Any) -> StageResult:
        """test_cmd 실행."""
        start = datetime.utcnow()
        test_cmd = getattr(project, "test_cmd", None)
        if not test_cmd:
            return StageResult(PipelineStage.TEST, True, "테스트 명령어 미설정 (생략)", duration=_elapsed(start))
        try:
            from core.ssh_exec import execute_ssh_command
            server_id = getattr(project, "server_id", None)
            result = await execute_ssh_command(server_id, test_cmd, timeout=_DEPLOY_TIMEOUT)
            ok = result.exit_code == 0
            return StageResult(
                PipelineStage.TEST, ok,
                "테스트 통과" if ok else f"테스트 실패 (rc={result.exit_code})",
                stdout=result.stdout[-3000:],
                duration=_elapsed(start),
            )
        except Exception as exc:
            return StageResult(PipelineStage.TEST, False, str(exc), duration=_elapsed(start))

    async def _backup(self, session_id: int, project: Any) -> StageResult:
        """현재 배포본 백업."""
        start = datetime.utcnow()
        deploy_path = getattr(project, "deploy_path", None)
        if not deploy_path:
            return StageResult(PipelineStage.BACKUP, True, "deploy_path 미설정 (백업 생략)", duration=_elapsed(start))
        try:
            from core.ssh_exec import execute_ssh_command
            server_id = getattr(project, "server_id", None)
            cmd = f"cp -rf {deploy_path} {deploy_path}.backup 2>/dev/null && echo OK"
            result = await execute_ssh_command(server_id, cmd, timeout=300)
            ok = "OK" in (result.stdout or "")
            return StageResult(PipelineStage.BACKUP, ok, "백업 완료" if ok else "백업 실패 (경고)", duration=_elapsed(start))
        except Exception as exc:
            return StageResult(PipelineStage.BACKUP, False, f"백업 오류: {exc}", duration=_elapsed(start))

    async def _deploy(self, session_id: int, project: Any) -> StageResult:
        """SSH로 파일 전송 및 deploy_path 배포."""
        start = datetime.utcnow()
        deploy_path = getattr(project, "deploy_path", None)
        deploy_cmd  = getattr(project, "deploy_cmd", None)
        if not deploy_path and not deploy_cmd:
            return StageResult(PipelineStage.DEPLOY, True, "배포 설정 미지정 (생략)", duration=_elapsed(start))
        try:
            from core.ssh_exec import execute_ssh_command
            server_id = getattr(project, "server_id", None)
            cmd = deploy_cmd or f"ls {deploy_path} > /dev/null && echo DEPLOYED"
            result = await execute_ssh_command(server_id, cmd, timeout=_DEPLOY_TIMEOUT)
            ok = result.exit_code == 0
            return StageResult(
                PipelineStage.DEPLOY, ok,
                "배포 완료" if ok else f"배포 실패 (rc={result.exit_code})",
                stdout=result.stdout[-2000:],
                duration=_elapsed(start),
            )
        except Exception as exc:
            return StageResult(PipelineStage.DEPLOY, False, str(exc), duration=_elapsed(start))

    async def _restart(self, session_id: int, project: Any) -> StageResult:
        """WAS/서비스 재시작."""
        start = datetime.utcnow()
        restart_cmd = getattr(project, "was_restart_cmd", None)
        if not restart_cmd:
            return StageResult(PipelineStage.RESTART, True, "재시작 명령어 미설정 (생략)", duration=_elapsed(start))
        try:
            from core.ssh_exec import execute_ssh_command
            server_id = getattr(project, "server_id", None)
            result = await execute_ssh_command(server_id, restart_cmd, timeout=120)
            ok = result.exit_code == 0
            return StageResult(
                PipelineStage.RESTART, ok,
                "서비스 재시작 완료" if ok else f"재시작 실패 (rc={result.exit_code})",
                duration=_elapsed(start),
            )
        except Exception as exc:
            return StageResult(PipelineStage.RESTART, False, str(exc), duration=_elapsed(start))

    async def _health_check(self, session_id: int, project: Any) -> StageResult:
        """HTTP GET 헬스체크 (최대 _HEALTH_CHECK_RETRIES 회 재시도)."""
        start = datetime.utcnow()
        health_url = getattr(project, "health_check_url", None)
        if not health_url:
            return StageResult(PipelineStage.HEALTH_CHECK, True, "헬스체크 URL 미설정 (생략)", duration=_elapsed(start))

        for attempt in range(1, _HEALTH_CHECK_RETRIES + 1):
            try:
                async with httpx.AsyncClient(timeout=10) as client:
                    resp = await client.get(health_url)
                if resp.status_code < 400:
                    return StageResult(
                        PipelineStage.HEALTH_CHECK, True,
                        f"헬스체크 통과 (시도 {attempt}/{_HEALTH_CHECK_RETRIES}, HTTP {resp.status_code})",
                        duration=_elapsed(start),
                    )
                logger.warning("헬스체크 실패: attempt=%d/%d status=%d", attempt, _HEALTH_CHECK_RETRIES, resp.status_code)
            except Exception as exc:
                logger.warning("헬스체크 오류: attempt=%d/%d err=%s", attempt, _HEALTH_CHECK_RETRIES, exc)

            if attempt < _HEALTH_CHECK_RETRIES:
                await asyncio.sleep(_HEALTH_CHECK_INTERVAL)

        return StageResult(
            PipelineStage.HEALTH_CHECK, False,
            f"헬스체크 {_HEALTH_CHECK_RETRIES}회 모두 실패",
            duration=_elapsed(start),
        )

    async def _notify(self, session_id: int, success: bool, message: str) -> StageResult:
        """WorkLog 등록 + 메신저 알림."""
        start = datetime.utcnow()
        try:
            from database import SessionLocal
            from models import VibeSession, WorkLog
            from core.notify import send_messenger
            import os

            async with SessionLocal() as db:
                session = await db.get(VibeSession, session_id)
                log = WorkLog(
                    sr_id       = session.sr_id if session else None,
                    work_type   = "DEPLOY",
                    content     = message,
                    result_status = "PASS" if success else "FAIL",
                    created_by  = "deploy_pipeline",
                )
                db.add(log)
                await db.commit()

            icon = "✅" if success else "❌"
            room = os.getenv("MESSENGER_OPS_ROOM", "ops")
            await send_messenger(room, {
                "type": "text",
                "text": f"{icon} [배포 알림] 세션 {session_id}\n{message}",
            })

            return StageResult(PipelineStage.NOTIFY, True, "알림 발송 완료", duration=_elapsed(start))
        except Exception as exc:
            logger.warning("알림 발송 실패: %s", exc)
            return StageResult(PipelineStage.NOTIFY, True, f"알림 발송 실패 (무시): {exc}", duration=_elapsed(start))

    # ── 내부 유틸 ─────────────────────────────────────────────────────────────

    def _advance(self, session_id: int, next_stage: PipelineStage, prev_result: StageResult) -> None:
        """파이프라인 진행 단계를 업데이트한다."""
        status = self._running.get(session_id)
        if status is None:
            return
        if prev_result.success:
            status.completed_stages.append(prev_result.stage.value)
        status.current_stage = next_stage

    async def _update_session(
        self,
        session_id: int,
        status: str,
        message: str,
        mark_built: bool = False,
        mark_tested: bool = False,
        mark_deployed: bool = False,
    ) -> None:
        """tb_vibe_session 상태 업데이트."""
        from database import SessionLocal
        from models import VibeSession
        now = datetime.utcnow()
        try:
            async with SessionLocal() as db:
                session = await db.get(VibeSession, session_id)
                if session:
                    session.status    = status
                    session.error_msg = message if "FAILED" in status else None
                    if mark_built:   session.built_at    = now
                    if mark_tested:  session.tested_at   = now
                    if mark_deployed: session.deployed_at = now
                    await db.commit()
        except Exception as exc:
            logger.warning("세션 상태 업데이트 실패: session_id=%d err=%s", session_id, exc)


# ── 유틸 ─────────────────────────────────────────────────────────────────────

def _elapsed(start: datetime) -> float:
    return (datetime.utcnow() - start).total_seconds()


# ── 싱글턴 ───────────────────────────────────────────────────────────────────

_pipeline_instance: Optional[DeployPipeline] = None


def get_deploy_pipeline() -> DeployPipeline:
    """싱글턴 DeployPipeline 인스턴스를 반환한다."""
    global _pipeline_instance
    if _pipeline_instance is None:
        _pipeline_instance = DeployPipeline()
    return _pipeline_instance