26J30k3
"""
step6 Excel Builder v2 — KST 보정 + 차트 20개 + 워크로드별 분석 시트 + [파일명 res_usage_ 및 devops-test 버킷 반영]
"""
import os
import boto3
import pandas as pd
import numpy as np
from pathlib import Path
from openpyxl import Workbook
from openpyxl.styles import Font, PatternFill, Alignment, Border, Side
from openpyxl.utils import get_column_letter
from openpyxl.drawing.image import Image as XLImage
from openpyxl.formatting.rule import ColorScaleRule, DataBarRule
# ─── 📂 디렉토리 경로 ./data 기준 표준화 ──────────────────────────
BASE_DATA_DIR = Path("./data")
MERGED_DIR = BASE_DATA_DIR / "merged"
PLOT_DIR = BASE_DATA_DIR / "output" / "plots"
OUT_DIR = BASE_DATA_DIR / "output"
OUT_DIR.mkdir(parents=True, exist_ok=True)
C = dict(
hdr_dark="1F4E79", hdr_mid="2E75B6", hdr_light="BDD7EE",
accent="ED7D31", red="C00000", green="70AD47",
yellow="FFC000", gray_row="F2F2F2", white="FFFFFF",
border="9DC3E6", summary_bg="EBF3FB", purple="7030A0",
teal="00B0A0",
)
WL_COLORS_HEX = {
"SPARK_EXECUTOR":"1F4E79","SPARK_DRIVER":"2E75B6",
"AIRFLOW_WORKER":"ED7D31","AIRFLOW_SCHEDULER":"FFC000",
"STARROCKS_BE":"70AD47","STARROCKS_FE":"375623",
"JUPYTERLAB":"7030A0","GENERAL_APPS":"888888","POSTGRESQL":"C00000",
}
def ft(bold=False, size=10, color="000000", name="Arial"):
return Font(name=name, bold=bold, size=size, color=color)
def fill(hex_color):
return PatternFill("solid", fgColor=hex_color)
def thin_border():
t = Side(style="thin", color=C["border"])
return Border(left=t, right=t, top=t, bottom=t)
def center(wrap=False):
return Alignment(horizontal="center", vertical="center", wrap_text=wrap)
def left(wrap=False):
return Alignment(horizontal="left", vertical="center", wrap_text=wrap)
def set_col_widths(ws, widths):
for col, w in widths.items():
ws.column_dimensions[col].width = w
def apply_header_row(ws, row_idx, headers, bg=None, fg=C["white"], size=10):
bg = bg or C["hdr_dark"]
for c, h in enumerate(headers, 1):
cell = ws.cell(row=row_idx, column=c, value=h)
cell.font = ft(bold=True, size=size, color=fg)
cell.fill = fill(bg)
cell.alignment = center(wrap=True)
cell.border = thin_border()
def apply_data_rows(ws, df, start_row, num_formats=None, status_col_idx=None, zebra=True):
nf = num_formats or {}
for r_offset, row in enumerate(df.itertuples(index=False)):
row_num = start_row + r_offset
bg_hex = C["gray_row"] if (r_offset % 2 == 1 and zebra) else C["white"]
for c_idx, val in enumerate(row, 1):
cell = ws.cell(row=row_num, column=c_idx, value=val)
cell.font = ft(size=9)
cell.fill = fill(bg_hex)
cell.alignment = left()
cell.border = thin_border()
if c_idx in nf:
cell.number_format = nf[c_idx]
if status_col_idx and c_idx == status_col_idx:
v = str(val)
if "OOM" in v or "Killed" in v:
cell.fill = fill("FFCCCC"); cell.font = ft(bold=True, size=9, color=C["red"])
elif "Shortage" in v or "부족" in v:
cell.fill = fill("FFF2CC"); cell.font = ft(bold=True, size=9, color="7F6000")
elif "Over" in v or "과다" in v:
cell.fill = fill("DDEEFF"); cell.font = ft(bold=True, size=9, color=C["hdr_dark"])
elif "Optim" in v or "최적" in v:
cell.fill = fill("E2EFDA"); cell.font = ft(bold=True, size=9, color="375623")
return start_row + len(df)
def freeze_and_filter(ws, row=2):
ws.freeze_panes = ws.cell(row=row+1, column=1)
ws.auto_filter.ref = ws.dimensions
def add_chart_image(ws, path_key, anchor_cell, w=860, h=400, label=None):
path = PLOT_DIR / path_key
if not path.exists():
print(f" ⚠️ [차트 유실] {path_key} 파일이 {PLOT_DIR}에 없어 삽입을 건너뜁니다.")
return
row_num = int(''.join(filter(str.isdigit, anchor_cell)))
if label:
col = anchor_cell[0]
ws.cell(row=row_num-1, column=1, value=label).font = ft(bold=True, size=11, color=C["hdr_dark"])
img = XLImage(str(path))
img.width = w; img.height = h
ws.add_image(img, anchor_cell)
print(f" -> 🎨 차트 맵핑 완료: {path_key} ➡️ {anchor_cell} 셀")
# ─── Sheet 0: 전사 종합 요약 ───────────────────────────────
def build_sheet_summary(wb, df_pod, df_ns, infra_tag):
print("⏳ [0/8] '0. 전사종합요약' 대시보드 탭 구축 개시...")
ws = wb.active
ws.title = "0. 전사종합요약"
ws.sheet_view.showGridLines = False
ws.row_dimensions[1].height = 42
ws.merge_cells("A1:H1")
t = ws["A1"]
t.value = f"Resource Governance Master Report [{infra_tag}] (KST 기준)"
t.font = ft(bold=True, size=15, color=C["white"])
t.fill = fill(C["hdr_dark"]); t.alignment = center()
oom_cnt = int(df_pod["is_oom_killed"].sum())
no_req_cnt = int((df_pod["has_no_request"] | df_pod["has_no_limit"]).sum())
alloc_ch = df_pod["cpu_allocated_core_hours"].sum()
usage_ch = df_pod["cpu_usage_core_hours"].sum()
waste_ch = df_pod["cpu_waste_core_hours"].sum()
eff_pct = usage_ch / max(alloc_ch, 0.001) * 100
mem_waste = df_pod["mem_waste_gb_hours"].sum()
top30 = max(1, int(len(df_pod)*0.30))
kst_dates = sorted(df_pod["date"].unique())
date_range = f"{kst_dates[0]} ~ {kst_dates[-1]} (KST)"
kpis = [
("정산 대상 인프라 도메인", infra_tag, "002060", C["white"]),
("분석 기간 (KST)", date_range, C["hdr_dark"], C["white"]),
("총 관측 컨테이너 수", f"{len(df_pod):,} 개", C["hdr_dark"], C["white"]),
("OOMKilled 발생 컨테이너", f"{oom_cnt:,} 개", C["red"], C["white"]),
("리소스 미설정 위반 컨테이너", f"{no_req_cnt:,} 개", "E26B0A", C["white"]),
("전사 CPU 낭비 총량", f"{waste_ch:,.1f} Core-H", C["hdr_mid"], C["white"]),
("전사 CPU 평균 활용률", f"{eff_pct:.1f} %", "375623", C["white"]),
("전사 Memory 낭비 총량", f"{mem_waste:,.1f} GB-H", "6B4F9B", C["white"]),
("최적화 권고 대상 (Top 30%)", f"{top30:,} 개", C["accent"], C["white"]),
]
ws.row_dimensions[2].height = 6
for i, (label, value, bg, fg) in enumerate(kpis):
row = 3 + i
ws.row_dimensions[row].height = 28
lc = ws.cell(row=row, column=1, value=label)
lc.font = ft(bold=True, size=10, color=C["hdr_dark"])
lc.fill = fill(C["summary_bg"]); lc.alignment = left(); lc.border = thin_border()
ws.merge_cells(f"A{row}:C{row}")
vc = ws.cell(row=row, column=4, value=value)
vc.font = ft(bold=True, size=11, color=fg)
vc.fill = fill(bg); vc.alignment = center(); vc.border = thin_border()
ws.merge_cells(f"D{row}:F{row}")
set_col_widths(ws, {"A":28,"B":14,"C":14,"D":22,"E":14,"F":14,"G":20,"H":20})
chart_row = 15
ws.cell(row=chart_row-1, column=1, value="[ 거버넌스 현황 분포 ]").font = ft(bold=True, size=11, color=C["hdr_dark"])
ws.cell(row=chart_row-1, column=5, value=f"[ {infra_tag} 네임스페이스 파레토 ]").font = ft(bold=True, size=11, color=C["hdr_dark"])
add_chart_image(ws, "chart6_status_donut.png", f"A{chart_row}", w=420, h=300)
add_chart_image(ws, "chart5_pareto_ns_waste.png", f"E{chart_row}", w=560, h=300)
chart_row2 = chart_row + 19
ws.cell(row=chart_row2-1, column=1, value="[ Waste Footprint Bubble ]").font = ft(bold=True, size=11, color=C["hdr_dark"])
ws.cell(row=chart_row2-1, column=5, value="[ Workload 파레토 분석 ]").font = ft(bold=True, size=11, color=C["hdr_dark"])
add_chart_image(ws, "chart7_waste_footprint_bubble.png", f"A{chart_row2}", w=480, h=340)
add_chart_image(ws, "chart11_pareto_workload_waste.png", f"E{chart_row2}", w=540, h=340)
# ─── Sheet 1: 파레토 분석 NS ──────────────────────────────
def build_sheet_pareto(wb, df_ns, infra_tag):
print(f"⏳ [1/8] '1. 파레토분석_NS' 시트 렌더링 중... (대상 테넌트 수: {len(df_ns)}개)")
ws = wb.create_sheet("1. 파레토분석_NS")
ws.sheet_view.showGridLines = False
ws.merge_cells("A1:I1")
t = ws["A1"]
t.value = f"Namespace별 CPU Waste 파레토 분석 [{infra_tag}]"
t.font = ft(bold=True, size=13, color=C["white"])
t.fill = fill(C["hdr_dark"]); t.alignment = center()
ws.row_dimensions[1].height = 32
headers = ["Namespace","실행시간 합계(분)","컨테이너 수","할당 Core-H","낭비 Core-H","낭비 비중(%)","누적 비중(%)","등급"]
apply_header_row(ws, 2, headers, bg=C["hdr_mid"])
df_disp = df_ns.copy()
df_disp["등급"] = df_disp["waste_cumsum_pct"].apply(
lambda x: "Critical (Top 20%)" if x<=20 else ("High (Top 50%)" if x<=50 else ("Medium" if x<=80 else "Low")))
col_order = ["namespace","minutes_running_sum","container_cnt","total_allocated_core_hours","total_waste_core_hours","waste_share_pct","waste_cumsum_pct","등급"]
df_out = df_disp[col_order]
nf = {4:"#,##0.0", 5:"#,##0.0", 6:"0.00", 7:"0.00"}
end_row = apply_data_rows(ws, df_out, start_row=3, num_formats=nf)
ws.conditional_formatting.add(
f"E3:E{end_row}",
DataBarRule(start_type="min", end_type="max", color="2E75B6", showValue=True))
set_col_widths(ws, {"A":22,"B":18,"C":14,"D":16,"E":16,"F":12,"G":12,"H":22})
freeze_and_filter(ws)
add_chart_image(ws, "chart5_pareto_ns_waste.png", f"A{end_row+3}", w=820, h=400,
label=f"[ {infra_tag} 네임스페이스 파레토 차트 ]")
# ─── Sheet 2: CPU 분석 ────────────────────────────────────
def build_sheet_cpu(wb, df_pod, infra_tag):
top30 = max(1, int(len(df_pod)*0.30))
print(f"⏳ [2/8] '2. CPU Request_Usage 분석' 시트 빌드 중... (상위 30% 격리: {top30}행)")
ws = wb.create_sheet("2. CPU Request_Usage 분석")
ws.sheet_view.showGridLines = False
ws.merge_cells("A1:L1")
t = ws["A1"]
t.value = f"CPU Resource Efficiency Analysis [{infra_tag}] — Request / Limit / Usage"
t.font = ft(bold=True, size=13, color=C["white"])
t.fill = fill(C["hdr_dark"]); t.alignment = center()
ws.row_dimensions[1].height = 32
headers = ["날짜(KST)","클러스터","네임스페이스","워크로드","Pod","컨테이너",
"CPU Request","CPU Limit","CPU P95","활용률(%)","낭비 Core-H","상태"]
apply_header_row(ws, 2, headers, bg=C["hdr_mid"])
df_out = df_pod.sort_values("cpu_waste_core_hours", ascending=False).head(top30).copy()
df_out["util"] = np.where(df_out["cpu_request_max"]>0,
(df_out["cpu_usage_p95"]/df_out["cpu_request_max"]*100).round(1), 0)
df_out["status_en"] = df_out["status"].map({
"💥 OOM장애발생":"OOM Killed","⚠️ Request부족":"Request Shortage",
"📉 과다할당":"Over-allocated","✅ 최적화완료":"Optimized"}).fillna("Unknown")
cols = ["date","cluster","namespace","workload_type","pod","container",
"cpu_request_max","cpu_limit_max","cpu_usage_p95","util","cpu_waste_core_hours","status_en"]
df_disp = df_out[cols].reset_index(drop=True)
nf = {7:"0.000",8:"0.000",9:"0.000",10:"0.0",11:"#,##0.0"}
end_row = apply_data_rows(ws, df_disp, start_row=3, num_formats=nf, status_col_idx=12)
ws.conditional_formatting.add(f"J3:J{end_row}",
ColorScaleRule(start_type="num", start_value=0, start_color="FF0000",
mid_type="num", mid_value=50, mid_color="FFFF00",
end_type="num", end_value=100, end_color="00B050"))
set_col_widths(ws, {"A":12,"B":18,"C":20,"D":18,"E":28,"F":16,"G":14,"H":14,"I":14,"J":14,"K":14,"L":16})
freeze_and_filter(ws)
add_chart_image(ws, "chart1_cpu_req_vs_usage_by_workload.png", f"A{end_row+3}", w=860, h=400, label="[ CPU Request / Limit / P95 by Workload ]")
add_chart_image(ws, "chart9_boxplot_cpu_util_by_workload.png", f"A{end_row+30}", w=860, h=400, label="[ CPU Utilization Boxplot by Workload ]")
# ─── Sheet 3: Memory 분석 ─────────────────────────────────
def build_sheet_memory(wb, df_pod, infra_tag):
top30 = max(1, int(len(df_pod)*0.30))
print(f"⏳ [3/8] '3. Memory Request_Usage 분석' 시트 빌드 중... (상위 30% 격리: {top30}행)")
ws = wb.create_sheet("3. Memory Request_Usage 분석")
ws.sheet_view.showGridLines = False
ws.merge_cells("A1:L1")
t = ws["A1"]
t.value = f"Memory Resource Efficiency Analysis [{infra_tag}] — Request / Limit / Usage"
t.font = ft(bold=True, size=13, color=C["white"])
t.fill = fill(C["hdr_dark"]); t.alignment = center()
ws.row_dimensions[1].height = 32
headers = ["날짜(KST)","클러스터","네임스페이스","워크로드","Pod","컨테이너",
"Mem Request(GB)","Mem Limit(GB)","Mem P95(GB)","활용률(%)","낭비 GB-H","상태"]
apply_header_row(ws, 2, headers, bg=C["hdr_mid"])
df_out = df_pod.sort_values("mem_waste_gb_hours", ascending=False).head(top30).copy()
df_out["util"] = np.where(df_out["mem_request_max"]>0,
(df_out["mem_usage_p95"]/df_out["mem_request_max"]*100).round(1), 0)
df_out["status_en"] = df_out["status"].map({
"💥 OOM장애발생":"OOM Killed","⚠️ Request부족":"Request Shortage",
"📉 과다할당":"Over-allocated","✅ 최적화완료":"Optimized"}).fillna("Unknown")
cols = ["date","cluster","namespace","workload_type","pod","container",
"mem_request_max","mem_limit_max","mem_usage_p95","util","mem_waste_gb_hours","status_en"]
df_disp = df_out[cols].reset_index(drop=True)
nf = {7:"0.000",8:"0.000",9:"0.000",10:"0.0",11:"#,##0.0"}
end_row = apply_data_rows(ws, df_disp, start_row=3, num_formats=nf, status_col_idx=12)
ws.conditional_formatting.add(f"J3:J{end_row}",
ColorScaleRule(start_type="num", start_value=0, start_color="FF0000",
mid_type="num", mid_value=50, mid_color="FFFF00",
end_type="num", end_value=100, end_color="00B050"))
set_col_widths(ws, {"A":12,"B":18,"C":20,"D":18,"E":28,"F":16,"G":14,"H":14,"I":14,"J":14,"K":14,"L":16})
freeze_and_filter(ws)
# ─── Sheet 4: OOM 및 자원 부족 ───────────────────────────
def build_sheet_oom(wb, df_pod, infra_tag):
print("⏳ [4/8] '4. 자원부족및OOM장애군' 리스크 단속 시트 작성 중...")
ws = wb.create_sheet("4. 자원부족및OOM장애군")
ws.sheet_view.showGridLines = False
ws.merge_cells("A1:K1")
t = ws["A1"]
t.value = f"OOMKilled / CPU Request 부족 컨테이너 명세 [{infra_tag}]"
t.font = ft(bold=True, size=13, color=C["white"])
t.fill = fill(C["red"]); t.alignment = center()
ws.row_dimensions[1].height = 32
headers = ["날짜(KST)","클러스터","네임스페이스","워크로드","Pod","컨테이너",
"상태","CPU Request","CPU P95","Mem Limit(GB)","Mem P95(GB)"]
apply_header_row(ws, 2, headers, bg=C["red"])
df_out = df_pod[
(df_pod["cpu_shortage_cores"]>0) | (df_pod["is_oom_killed"])
].sort_values(["is_oom_killed","cpu_shortage_cores"], ascending=[False,False]).copy()
print(f" -> 💥 감지된 가용성 장애 수: 총 {len(df_out)}건")
df_out["status_en"] = df_out["status"].map({
"💥 OOM장애발생":"OOM Killed","⚠️ Request부족":"Request Shortage",
"📉 과다할당":"Over-allocated","✅ 최적화완료":"Optimized"}).fillna("Unknown")
cols = ["date","cluster","namespace","workload_type","pod","container",
"status_en","cpu_request_max","cpu_usage_p95","mem_limit_max","mem_usage_p95"]
df_disp = df_out[cols].reset_index(drop=True)
nf = {8:"0.000",9:"0.000",10:"0.000",11:"0.000"}
end_row = apply_data_rows(ws, df_disp, start_row=3, num_formats=nf, status_col_idx=7)
set_col_widths(ws, {"A":12,"B":18,"C":20,"D":18,"E":28,"F":16,"G":16,"H":14,"I":14,"J":14,"K":14})
freeze_and_filter(ws)
# ─── Sheet 5: 리소스 미설정 위반 ─────────────────────────
def build_sheet_violations(wb, df_pod, infra_tag):
print("⏳ [5/8] '5. 리소스미설정위반군' 거버넌스 블랙리스트 추출 중...")
ws = wb.create_sheet("5. 리소스미설정위반군")
ws.sheet_view.showGridLines = False
ws.merge_cells("A1:L1")
t = ws["A1"]
t.value = f"Resource Request / Limit 미설정 위반 컨테이너 목록 [{infra_tag}]"
t.font = ft(bold=True, size=13, color=C["white"])
t.fill = fill("E26B0A"); t.alignment = center()
ws.row_dimensions[1].height = 32
headers = ["날짜(KST)","클러스터","네임스페이스","워크로드","Pod","컨테이너",
"Request 미설정","Limit 미설정","CPU Request","CPU Limit","Mem Request(GB)","Mem Limit(GB)"]
apply_header_row(ws, 2, headers, bg="E26B0A")
df_out = df_pod[df_pod["has_no_request"]|df_pod["has_no_limit"]].sort_values("minutes_running", ascending=False).copy()
print(f" -> ⚠️ 규격 미달 위반 팟 수량: {len(df_out)}건")
df_out["req_flag"] = df_out["has_no_request"].map({True:"MISSING",False:"OK"})
df_out["lim_flag"] = df_out["has_no_limit"].map({True:"MISSING",False:"OK"})
cols = ["date","cluster","namespace","workload_type","pod","container",
"req_flag","lim_flag","cpu_request_max","cpu_limit_max","mem_request_max","mem_limit_max"]
df_disp = df_out[cols].reset_index(drop=True)
nf = {9:"0.000",10:"0.000",11:"0.000",12:"0.000"}
end_row = apply_data_rows(ws, df_disp, start_row=3, num_formats=nf)
set_col_widths(ws, {"A":12,"B":18,"C":20,"D":18,"E":28,"F":16,"G":14,"H":14,"I":14,"J":14,"K":14,"L":14})
freeze_and_filter(ws)
# ─── Sheet 6: 일별 트렌드 (KST) ──────────────────────────
def build_sheet_trends(wb, df_pod, infra_tag):
print("⏳ [6/8] '6. 일별트렌드_KST' 시계열 파티션 압축 롤업 중...")
ws = wb.create_sheet("6. 일별트렌드_KST")
ws.sheet_view.showGridLines = False
ws.merge_cells("A1:J1")
t = ws["A1"]
t.value = f"Daily Resource Trend [{infra_tag}] — KST 기준"
t.font = ft(bold=True, size=13, color=C["white"])
t.fill = fill(C["hdr_dark"]); t.alignment = center()
ws.row_dimensions[1].height = 32
note_cell = ws.cell(row=2, column=1,
value=f"※ 본 트렌드는 [{infra_tag}] 데이터의 KST(UTC+9) 타임라인 보정 집계 결과입니다.")
note_cell.font = ft(size=9, color="7F7F7F", bold=False)
ws.merge_cells("A2:J2")
df_daily = df_pod.groupby("date").agg(
containers=("container","count"),
cpu_alloc=("cpu_allocated_core_hours","sum"),
cpu_used=("cpu_usage_core_hours","sum"),
cpu_waste=("cpu_waste_core_hours","sum"),
mem_alloc=("mem_allocated_gb_hours","sum"),
mem_waste=("mem_waste_gb_hours","sum"),
oom_cnt=("is_oom_killed","sum"),
shortage_cnt=("cpu_shortage_cores", lambda x: (x>0).sum()),
).reset_index()
df_daily["cpu_util_pct"] = (df_daily["cpu_used"]/df_daily["cpu_alloc"].clip(lower=0.001)*100).round(1)
df_daily["mem_util_pct"] = ((df_daily["mem_alloc"]-df_daily["mem_waste"])/df_daily["mem_alloc"].clip(lower=0.001)*100).round(1)
headers = ["날짜(KST)","컨테이너","CPU 할당 Core-H","CPU 사용 Core-H","CPU 낭비 Core-H",
"Mem 할당 GB-H","Mem 낭비 GB-H","OOM 발생","Request 부족","CPU 활용률(%)","Mem 활용률(%)"]
apply_header_row(ws, 3, headers, bg=C["hdr_mid"])
nf = {3:"#,##0.0",4:"#,##0.0",5:"#,##0.0",6:"#,##0.0",7:"#,##0.0",8:"#,##0",9:"#,##0",10:"0.0",11:"0.0"}
df_disp = df_daily[["date","containers","cpu_alloc","cpu_used","cpu_waste",
"mem_alloc","mem_waste","oom_cnt","shortage_cnt","cpu_util_pct","mem_util_pct"]]
end_row = apply_data_rows(ws, df_disp, start_row=4, num_formats=nf)
for col_letter, label in [("J","CPU"), ("K","Mem")]:
ws.conditional_formatting.add(
f"{col_letter}4:{col_letter}{end_row}",
ColorScaleRule(start_type="num",start_value=0, start_color="FF0000", mid_type="num", mid_value=50, mid_color="FFFF00", end_type="num", end_value=100, end_color="00B050"))
set_col_widths(ws, {"A":14,"B":12,"C":16,"D":16,"E":16,"F":16,"G":14,"H":12,"I":14,"J":14,"K":14})
ws.freeze_panes = "A5"
for chart_key, anchor, w_img, h_img, lbl in [
("chart3_daily_waste_stack.png", f"A{end_row+3}", 860, 400, f"[ Daily CPU Waste Stacked ({infra_tag}) ]"),
("chart4_cpu_efficiency_heatmap.png", f"A{end_row+30}", 860, 400, "[ CPU Utilization Heatmap ]"),
]:
add_chart_image(ws, chart_key, anchor, w=w_img, h=h_img, label=lbl)
# ─── Sheet 7: 워크로드별 심층 분석 ─────────────────────
def build_sheet_workload(wb, df_pod, infra_tag):
print("⏳ [7/8] '7. 워크로드별_심층분석' 오픈소스 스택별 다차원 집계 중...")
ws = wb.create_sheet("7. 워크로드별_심층분석")
ws.sheet_view.showGridLines = False
ws.merge_cells("A1:K1")
t = ws["A1"]
t.value = f"Workload Type별 심층 자원 효율화 분석 [{infra_tag}]"
t.font = ft(bold=True, size=13, color=C["white"])
t.fill = fill(C["purple"]); t.alignment = center()
ws.row_dimensions[1].height = 32
df_wl = df_pod.groupby("workload_type").agg(
containers=("container","count"),
cpu_req_avg=("cpu_request_max","mean"),
cpu_lim_avg=("cpu_limit_max","mean"),
cpu_p95_avg=("cpu_usage_p95","mean"),
cpu_waste_sum=("cpu_waste_core_hours","sum"),
mem_req_avg=("mem_request_max","mean"),
mem_p95_avg=("mem_usage_p95","mean"),
mem_waste_sum=("mem_waste_gb_hours","sum"),
oom_cnt=("is_oom_killed","sum"),
shortage_cnt=("cpu_shortage_cores", lambda x: (x>0).sum()),
).reset_index()
df_wl["cpu_util_pct"] = (df_wl["cpu_p95_avg"]/df_wl["cpu_req_avg"].replace(0,np.nan)*100).round(1).fillna(0)
df_wl["mem_util_pct"] = (df_wl["mem_p95_avg"]/df_wl["mem_req_avg"].replace(0,np.nan)*100).round(1).fillna(0)
df_wl["lim_req_ratio"] = (df_wl["cpu_lim_avg"]/df_wl["cpu_req_avg"].replace(0,np.nan)).round(2).fillna(0)
df_wl = df_wl.sort_values("cpu_waste_sum", ascending=False).reset_index(drop=True)
headers2 = ["워크로드 타입","컨테이너 수","CPU Request(avg)","CPU Limit(avg)","CPU P95(avg)",
"CPU 활용률(%)","CPU 낭비 Core-H","Limit/Req 배율",
"Mem P95(avg GB)","Mem 활용률(%)","OOM 건수","CPU 부족 건수"]
apply_header_row(ws, 3, headers2, bg=C["purple"])
ws.cell(row=2, column=1, value=f"[ {infra_tag} 기술 스택별 집계 요약 ]").font = ft(bold=True, size=11, color=C["purple"])
cols_out = ["workload_type","containers","cpu_req_avg","cpu_lim_avg","cpu_p95_avg",
"cpu_util_pct","cpu_waste_sum","lim_req_ratio",
"mem_p95_avg","mem_util_pct","oom_cnt","shortage_cnt"]
df_disp = df_wl[cols_out]
nf = {3:"0.000",4:"0.000",5:"0.000",6:"0.0",7:"#,##0.0",8:"0.00",9:"0.000",10:"0.0"}
end_row = apply_data_rows(ws, df_disp, start_row=4, num_formats=nf)
for col_letter in ["F","J"]:
ws.conditional_formatting.add(f"{col_letter}4:{col_letter}{end_row}",
ColorScaleRule(start_type="num",start_value=0, start_color="FF0000", mid_type="num", mid_value=50, mid_color="FFFF00", end_type="num", end_value=100, end_color="00B050"))
set_col_widths(ws, {"A":20,"B":12,"C":14,"D":14,"E":14,"F":14,"G":14,"H":14,"I":14,"J":14,"K":12,"L":14})
ws.freeze_panes = "A5"
for chart_key, anchor, w_img, h_img, lbl in [
("chart15_oom_status_by_workload.png", f"A{end_row+3}", 860, 400, "[ Status Distribution by Workload ]"),
("chart9_boxplot_cpu_util_by_workload.png", f"A{end_row+30}", 860, 400, "[ CPU Utilization Boxplot ]"),
]:
add_chart_image(ws, chart_key, anchor, w=w_img, h=h_img, label=lbl)
# ─── Main ─────────────────────────────────────────────────
def main():
print("🚀 [Step6 개시] res_usage_ 계열 엑셀 리포트 빌더 v2 자동화 엔진 구동...")
p1 = MERGED_DIR / "enriched_fixed_7d.parquet"
p2 = MERGED_DIR / "pareto_fixed_ns.parquet"
print(f"📖 가공 원부 스캔 경로:\n - {p1}\n - {p2}")
if not p1.exists() or not p2.exists():
print("❌ 오류: step2 파이프라인의 가공 산출물(Parquet)이 디렉토리에 없습니다. step2를 먼저 실행하세요.")
return
df_pod = pd.read_parquet(p1)
df_ns = pd.read_parquet(p2)
print(f"✅ 데이터 로드 성료 -> Pod 명세: {len(df_pod):,}행 / Namespace 명세: {len(df_ns):,}행")
# ─── ⏱️ [요청 반영] 접두사를 finops_ 대신 res_usage_ 로 전면 변경 ───
meta_file = MERGED_DIR / "meta_run_info.txt"
if meta_file.exists():
with open(meta_file, "r") as f:
meta_str = f.read().strip()
infra_tag, date_tag = meta_str.split("@")
excel_name = f"res_usage_{infra_tag.lower()}_{date_tag}.xlsx"
print(f"🎯 런타임 메타데이터 감지 성료 -> 인프라: {infra_tag} | 일자: {date_tag}")
else:
infra_tag, date_tag = "ALL", "unknown"
excel_name = "res_usage_resource_governance_v2.xlsx"
print("⚠️ [주의] 런타임 메타데이터가 존재하지 않아 기본 파일명으로 양식을 고정합니다.")
wb = Workbook()
build_sheet_summary(wb, df_pod, df_ns, infra_tag)
build_sheet_pareto(wb, df_ns, infra_tag)
build_sheet_cpu(wb, df_pod, infra_tag)
build_sheet_memory(wb, df_pod, infra_tag)
build_sheet_oom(wb, df_pod, infra_tag)
build_sheet_violations(wb, df_pod, infra_tag)
build_sheet_trends(wb, df_pod, infra_tag)
build_sheet_workload(wb, df_pod, infra_tag)
out_path = OUT_DIR / excel_name
print(f"💾 메모리 스트림 디스크 쓰기 집행 중... ➡️ {out_path}")
wb.save(out_path)
print(f"📦 [로컬 마감 성료] 파일명: {excel_name} ({out_path.stat().st_size/1024:.0f} KB)\n")
# ─── 🪣 [요청 반영] 사내 MinIO AIStor devops-test 배포 버킷 레이어 ───
minio_endpoint = os.getenv("MINIO_ENDPOINT")
minio_access_key = os.getenv("MINIO_ACCESS_KEY")
minio_secret_key = os.getenv("MINIO_SECRET_KEY")
# 기본 폴백 대상 버킷명을 devops-test 로 전면 교체
bucket_name = os.getenv("MINIO_REPORT_BUCKET", "devops-test")
if all([minio_endpoint, minio_access_key, minio_secret_key]):
try:
s3_client = boto3.client(
"s3", endpoint_url=minio_endpoint,
aws_access_key_id=minio_access_key, aws_secret_access_key=minio_secret_key,
config=boto3.session.Config(signature_version="s3v4")
)
object_key = f"reports/{infra_tag.lower()}/{excel_name}"
print(f"🪣 [오브젝트 스토리지 싱크] devops-test 버킷 ➡️ S3://{object_key} 업로드 전송 중...")
s3_client.upload_file(str(out_path), bucket_name, object_key)
print("🏁 === [전사 배포 마감 성공] 최종 마스터 거버넌스 엑셀 자산 배포가 완수되었습니다. ===")
except Exception as e:
print(f"❌ [배포 오류] 사내 MinIO AIStor 업로드 중 예외 발생: {str(e)}")
else:
print("⚠️ [안내] MinIO 접속 환경변수가 생략되어 있어 원격 데이터 레이크 전송을 생략하고 로컬에서 마감합니다.")
if __name__ == "__main__":
main()
engineer