πŸ“ [Wiki] Dr.Kube β†’ AI-Powered Self-Diagnostic Agent for Kubernetes

Open Project
, , ,
1

:bookmark_tabs: [Dr.Kube] Official Wiki

KR: 이 νŽ˜μ΄μ§€λŠ” Dr.Kube의 λΉ„μ „, 기술적 λ°©ν–₯μ„±, 그리고 ν˜‘μ—… 방식을 μ •μ˜ν•˜λŠ” 톡합 λ¬Έμ„œμž…λ‹ˆλ‹€. νŒ€μ›κ³Ό μ™ΈλΆ€ κΈ°μ—¬μžλ“€μ΄ μ‘°ν™”λ‘­κ²Œ ν˜‘μ—…ν•  수 μžˆλ„λ‘ λ•λŠ” 곡식 κ°€μ΄λ“œλΌμΈμž…λ‹ˆλ‹€.

EN: This page serves as the comprehensive documentation defining the vision, technical direction, and collaboration methods for Dr.Kube. It is an official guideline to ensure seamless collaboration between the team and external contributors.

1. ν”„λ‘œμ νŠΈ κ°œμš” (Project Overview)

  • Purpose: ν΄λŸ¬μŠ€ν„°μ˜ 'λ‹€μž‰ λ©”μ‹œμ§€β€™μ„ 뢄석해 슀슀둜 μž₯μ• λ₯Ό μ§„λ‹¨ν•˜κ³  ν•΄κ²° κ°€μ΄λ“œκΉŒμ§€ μ œκ³΅ν•˜λŠ” 지λŠ₯ν˜• AI μ—μ΄μ „νŠΈ ; An autonomous AI agent that analyzes β€˜dying messages’ from failing clusters to diagnose root causes and deliver actionable remediation guidelines.

  • Background / Introduction (KR):

  • Kubernetes의 고질적인 문제인 μž₯μ•  λ°œμƒ μ‹œ MTTR(평균 볡ꡬ μ‹œκ°„)을 λ‹¨μΆ•ν•˜λŠ” 것을 λͺ©ν‘œλ‘œ ν•©λ‹ˆλ‹€.

  • LangGraph 기반 μΆ”λ‘  루프λ₯Ό 톡해 κ·Όλ³Έ 원인을 νƒμƒ‰ν•˜κ³ , Slack λ“± λ©”μ‹ μ €λ‘œ μš΄μ˜μžμ—κ²Œ 즉각적인 쑰치 κ°€μ΄λ“œλΌμΈμ„ μ œκ³΅ν•©λ‹ˆλ‹€.

  • 운영자의 승인 ν•˜μ— λͺ…령을 μˆ˜ν–‰ν•˜μ—¬ μ•ˆμ „μ„±μ„ ν™•λ³΄ν•˜λŠ” Safe AI Ops μƒνƒœκ³„λ₯Ό λ§Œλ“­λ‹ˆλ‹€.

  • Background / Introduction (EN):

  • Our objective is to minimize Mean Time to Recovery (MTTR), a persistent challenge in Kubernetes environments during system failures. By leveraging LangGraph-based reasoning loops, the system identifies root causes and delivers immediate remediation guidelines to operators via messaging platforms like Slack. We are building a β€˜Safe AIOps’ ecosystem that ensures operational integrity by executing commands only upon explicit human-in-the-loop approval.

  • Core Values:

  • Safe AIOps (μ•ˆμ „ν•œ AI 운영) β€” Human-in-the-Loop 승인 기반의 μ•ˆμ „ν•œ μžλ™ 볡ꡬ / Safe automated recovery based on human-in-the-loop approval

  • GitOps First (GitOps μš°μ„ ) β€” λͺ¨λ“  변경은 Git을 ν†΅ν•΄μ„œλ§Œ, ν΄λŸ¬μŠ€ν„° 직접 μˆ˜μ • κΈˆμ§€ / All changes through Git only, no direct cluster modifications

  • Observability (κ΄€μΈ‘ κ°€λŠ₯μ„±) β€” 4가지 μ‹œκ·Έλ„(Metrics, Logs, Traces, Profiles) 기반 κ·Όλ³Έ 원인 뢄석 / Root cause analysis based on 4 signals (Metrics, Logs, Traces, Profiles)

2. νŒ€ ꡬ성 (The Team)

Roles and responsibilities for the member team.

|이름 (Name)|ID|μ—­ν•  (Role)|SNS|μ£Όμš” μ±…μž„ (Responsibilities - KR/EN)|

| β€” | β€” | β€” | β€” | β€” |

|λ°±μ’…ν™”|@jonghwa|Team Leader|LinkedIn|λ‘œλ“œλ§΅ 및 μ΅œμ’… μ˜μ‚¬κ²°μ • / Roadmap & Final decision-making|

|κΉ€νƒœλΉˆ|@taebin|Member|LinkedIn|멀버 / Member|

|λ°•μŠΉκ·œ|@seunggyu|Member|LinkedIn|멀버 / Member|

|μœ μ§„μŠΉ|@jinseung|Member|LinkedIn|멀버 / Member|

|μž„μž¬ν›ˆ|@jaehoon|Member|LinkedIn|멀버 / Member|

3. 기술 μŠ€νƒ (Tech Stack)

  • Language: Python 3.11+

  • Framework: LangGraph, FastAPI

  • LLM: Google Gemini Flash, Ollama (local fallback)

  • Infra: Kubernetes (Kind), ArgoCD, Docker, Helm, Chaos Mesh

  • Observability: Prometheus, Grafana, Loki, Tempo, Pyroscope, Alloy

  • Security: SOPS + age, cert-manager (Let’s Encrypt)

  • Communication: Slack, GitHub Issues, Discord

4. λ‘œλ“œλ§΅ (Roadmap)

  • Phase 1: κ΄€μΈ‘μ„± μŠ€νƒ ꡬ좕 및 μ•Œλ¦Ό 체계 μ™„μ„± (Observability Stack & Alert System) :white_check_mark:

  • Phase 2: LangGraph μ—μ΄μ „νŠΈ μ›Œν¬ν”Œλ‘œμš° 개발 및 카였슀 ν…ŒμŠ€νŠΈ (Agent Workflow Dev & Chaos Testing) :hourglass_not_done:

  • Phase 3: Human-in-the-Loop ν”Όλ“œλ°± 루프 및 E2E 톡합 ν…ŒμŠ€νŠΈ (HITL Feedback & E2E Integration Test)

5. μ°Έμ—¬ 방법 (How to Contribute)

  • Branch Strategy / 브랜치 μ „λž΅:

  • main λΈŒλžœμΉ˜κ°€ 기본이며, ArgoCDκ°€ μžλ™ λ™κΈ°ν™”ν•©λ‹ˆλ‹€. / main branch is the default; ArgoCD auto-syncs from it.

  • μž‘μ—… μ‹œ feature 브랜치λ₯Ό μƒμ„±ν•˜κ³  PR을 톡해 λ³‘ν•©ν•©λ‹ˆλ‹€. / Create feature branches for work and merge via PR.

  • Issues: λ²„κ·Έλ‚˜ κΈ°λŠ₯ μ œμ•ˆμ€ GitHub Issuesλ₯Ό ν™œμš©ν•˜μ„Έμš”. / Please use GitHub Issues for bug reports or feature requests.

  • PRs: λͺ¨λ“  Pull RequestλŠ” 리더(@b100to)의 리뷰 ν›„ λ³‘ν•©λ©λ‹ˆλ‹€. / All PRs require review by the Team Leader (@b100to) before merging.

  • GitOps 원칙 / GitOps Principles:

  • kubectl apply/patch λ“± ν΄λŸ¬μŠ€ν„° 직접 μˆ˜μ • κΈˆμ§€ β€” 변경은 였직 Git을 ν†΅ν•΄μ„œλ§Œ μˆ˜ν–‰ / No direct cluster modifications β€” all changes through Git only

  • values/*.yaml μˆ˜μ • β†’ PR 생성 β†’ ArgoCD Sync / Edit values/*.yaml β†’ Create PR β†’ ArgoCD Sync

  • Guide: CONTRIBUTING.md νŒŒμΌμ„ μ°Έκ³ ν•˜μ„Έμš”. / Please refer to the CONTRIBUTING.md file.

  • Discord (Official): [Dr.Kube Invite Link]

  • KR: μ‹€μ‹œκ°„ μ†Œν†΅ 및 기술 지원을 μœ„ν•œ μ±„λ„μž…λ‹ˆλ‹€.

  • EN: Official channel for real-time communication and technical support.

6. λ¦¬μ†ŒμŠ€ 및 링크 (Resources & Links)

| This is a space where knowledge is not merely consumed, but respected, sovereign, and connectedβ€”shared together with cloud industry professionals (Bros).|

| 지식이 μ†ŒλΉ„λ˜μ§€ μ•Šκ³  μ‘΄μ€‘Β·μ£ΌκΆŒλ³΄μž₯Β·μ—°κ²°λ˜λŠ” κ³΅κ°„μœΌλ‘œ ν΄λΌμš°λ“œ ν˜„μ—… μ „λ¬Έκ°€(Bro)와 ν•¨κ»˜ κ³΅μœ ν•˜κ³  μžˆμŠ΅λ‹ˆλ‹€. |

Open Project Season 2 쀑간 λ°œν‘œ - Dr.Kube 문의