diff --git a/README.md b/README.md index c0f4da7..53710cd 100644 --- a/README.md +++ b/README.md @@ -5,6 +5,36 @@ Advanced analysis tool for identifying optimal Ceph OSD replacement candidates based on multiple health, capacity, resilience, and performance factors. +![analyzeOSDs ranking Ceph OSD replacement candidates](docs/img/demo.gif) + +## ⚡ Quick run + +Run on any Ceph node as root (read-only: it only queries `ceph`, `smartctl` and `lsblk`, locally and over SSH). + +```bash +# Most common: HDD OSDs, with debug output +sudo python3 -c "import urllib.request; exec(urllib.request.urlopen('https://code.lotusguild.org/LotusGuild/analyzeOSDs/raw/branch/main/ceph_osd_analyzer.py').read().decode())" --debug --class hdd + +# All OSDs +sudo python3 -c "import urllib.request; exec(urllib.request.urlopen('https://code.lotusguild.org/LotusGuild/analyzeOSDs/raw/branch/main/ceph_osd_analyzer.py').read().decode())" + +# Only NVMe, or only drives of at least 8 TB +sudo python3 -c "import urllib.request; exec(urllib.request.urlopen('https://code.lotusguild.org/LotusGuild/analyzeOSDs/raw/branch/main/ceph_osd_analyzer.py').read().decode())" --class nvme +sudo python3 -c "import urllib.request; exec(urllib.request.urlopen('https://code.lotusguild.org/LotusGuild/analyzeOSDs/raw/branch/main/ceph_osd_analyzer.py').read().decode())" --min-size 8 +``` + +## Why this exists + +With 30+ OSDs of mixed ages and sizes, "which drive do I replace next?" is a judgement call that involves SMART health, drive age, how full the OSD is, how much capacity the swap would add, and how balanced the hosts are. This tool gathers all of it from the live cluster and turns it into a **ranked list with the reasons spelled out**, so the next purchase and the next maintenance window are chosen on evidence. + +## Gallery + +Ranked HDD replacement candidates on the live cluster. Each candidate shows a replacement score, health score and issues, and the capacity and balance effect of replacing it, followed by a per-host summary. (Here osd.2 ranks first because no SMART data could be read for it.) + +![analyzeOSDs --class hdd](docs/img/hdd-candidates.png) + +Rendered from real command output with `scripts/render_terminal.py`. + ## Overview This script provides comprehensive analysis of your Ceph cluster to identify which OSDs should be replaced first. It combines SMART health data, capacity optimization potential, cluster resilience impact, and performance metrics to generate a prioritized replacement list. diff --git a/docs/img/demo.gif b/docs/img/demo.gif new file mode 100644 index 0000000..a583d19 Binary files /dev/null and b/docs/img/demo.gif differ diff --git a/docs/img/hdd-candidates.png b/docs/img/hdd-candidates.png new file mode 100644 index 0000000..40b9a5a Binary files /dev/null and b/docs/img/hdd-candidates.png differ diff --git a/scripts/render_terminal.py b/scripts/render_terminal.py new file mode 100644 index 0000000..8ef4d95 --- /dev/null +++ b/scripts/render_terminal.py @@ -0,0 +1,104 @@ +"""Render CLI output as a styled terminal screenshot / animated GIF with Playwright + Pillow. + +Usage (as a module): + from term_render import render + render(title, command, lines, png='out.png', gif='out.gif', width=1000, rules=[(regex, css_color), ...]) +`lines` is a list of plain-text output lines. No real terminal is needed, so output can be sanitized first. +""" +import asyncio, html, io, re +from PIL import Image +from playwright.async_api import async_playwright + + +ANSI_COL={30:'#6e7681',31:'#ff5c6c',32:'#3ddc84',33:'#ffb454',34:'#58a6ff',35:'#d2a8ff',36:'#7fdbff',37:'#e6edf3', + 90:'#8b949e',91:'#ff7b86',92:'#56f09f',93:'#ffd479',94:'#79c0ff',95:'#e2c5ff',96:'#a5f3ff',97:'#ffffff'} +def ansi_to_html(line): + """Convert ANSI SGR colour/bold sequences to tags (everything else is stripped).""" + out=[]; open_spans=0; pos=0 + for m in re.finditer(r'\x1b\[([0-9;]*)m|\x1b\[[0-9;?]*[A-Za-z]',line): + out.append(html.escape(line[pos:m.start()])); pos=m.end() + if m.group(1) is None: continue + codes=[int(c) if c else 0 for c in m.group(1).split(';')] + for c in codes: + if c==0: out.append(''*open_spans); open_spans=0 + elif c==1: out.append(''); open_spans+=1 + elif c in ANSI_COL: out.append(f''); open_spans+=1 + out.append(html.escape(line[pos:])); out.append(''*open_spans) + return ''.join(out) + +def mask(lines, serial_cols=True, ips=True): + """Sanitize output before publishing: hide drive serials and private IPs.""" + res=[] + for l in lines: + if ips: l=re.sub(r'\b(10|192\.168|172\.(1[6-9]|2\d|3[01]))\.(\d+)\.(\d+)\.(\d+)\b','x.x.x.x',l) if not l.startswith('x') else l + if serial_cols and re.match(r'^(\d+|m2-\d)\s+/dev/',l): + parts=re.split(r'(\s{2,})',l) + # fields alternate with whitespace; serial is the 8th field (index 14 in the split list) + if len(parts)>14: + v=parts[14]; parts[14]=v[:2]+'\u2022'*max(0,len(v)-2) + l=''.join(parts) + res.append(l) + return res + +DEFAULT_RULES=[(r'✓|\bOK\b|\bNORMAL\b|\bPASS(ED)?\b|HEALTH_OK','#3ddc84'),(r'⚠️?|\bWARN(ING)?\b|HEALTH_WARN|\[WARN\]','#ffb454'),(r'\bERROR\b|\bFAIL(ED)?\b|\bCRITICAL\b|✗|HEALTH_ERR','#ff5c6c'),(r'^=+.*=+$|^┏.*|^┃.*|^┣.*|^┗.*','#7fdbff')] + +PAGE=''' + + +
%(t)s
''' + +def colorize(line, rules): + if '\x1b[' in line: return ansi_to_html(line) + s=html.escape(line) + for pat,col in rules: + if pat.startswith('^'): + if re.search(pat,line): return f'{s}' + else: + s=re.sub('('+pat+')',lambda m:f'{m.group(0)}',s) + return s + +async def _render(title,command,lines,png,gif,width,rules,min_h): + rules=rules or DEFAULT_RULES + prompt='root@node:~# ' + async with async_playwright() as p: + b=await p.chromium.launch(); pg=await b.new_page(viewport={'width':width+2,'height':600}) + await pg.set_content(PAGE%{'w':width,'h':min_h,'t':html.escape(title)}); await pg.wait_for_timeout(1200) + async def show(body,cursor=False): + await pg.evaluate("(h)=>{document.getElementById('t').innerHTML=h}",body+(' ' if cursor else '')) + async def shot(): + el=pg.locator('#win'); return await el.screenshot() + full=prompt+''+html.escape(command)+'\n'+'\n'.join(colorize(l,rules) for l in lines) + await show(full) + if png: open(png,'wb').write(await shot()); print('png',png) + if gif: + frames=[]; dur=[] + def add(img,d): frames.append(Image.open(io.BytesIO(img)).convert('RGB')); dur.append(d) + await show(prompt,True); add(await shot(),500) + for i in range(1,len(command)+1,3): + await show(prompt+''+html.escape(command[:i])+'',True); add(await shot(),45) + await show(prompt+''+html.escape(command)+'',True); add(await shot(),500) + head=prompt+''+html.escape(command)+'\n'; out=[] + step=max(1,len(lines)//20) + for i in range(0,len(lines),step): + out=lines[:i+step]; await show(head+'\n'.join(colorize(l,rules) for l in out)); add(await shot(),120) + await show(full); add(await shot(),3500) + w0=frames[0].width + scale=min(1.0,780/w0); frames=[f.resize((int(f.width*scale),int(f.height*scale)),Image.LANCZOS) for f in frames] + # all frames must share size: pad to the tallest + H=max(f.height for f in frames); W=frames[0].width + norm=[] + for f in frames: + c=Image.new('RGB',(W,H),(13,17,23)); c.paste(f,(0,0)); norm.append(c.quantize(colors=32,method=Image.MEDIANCUT,dither=Image.NONE)) + norm[0].save(gif,save_all=True,append_images=norm[1:],duration=dur,loop=0,optimize=True); print('gif',gif,len(norm)) + await b.close() + +def render(title,command,lines,png=None,gif=None,width=1000,rules=None,min_h=200): + asyncio.run(_render(title,command,lines,png,gif,width,rules,min_h))