Downloaded 18 times
Uploaded byShubhra Kar
Html5 devconf nodejs_devops_shubhra
The document outlines a presentation on using Node.js for API strategies, emphasizing its debugging, monitoring, and scaling capabilities. It highlights tools and solutions for diagnosing performance issues, managing memory, and ensuring production readiness. The speaker aims to convince the audience of Node.js's effectiveness and scalability for mobile applications and distributed systems.
Html5 devconf nodejs_devops_shubhra
- 1. Triage, Diagnose &Scale Node.js Shubhra Kar | Director – Products twitter:@shubhrakar mail:skar@strongloop.com
- 2. Why am Ihere…or rather who sent me here ?
- 3. Node Core Bert Belder Ben Noordhuis These guys sent me ! LoopBack/Express Core Raymond Feng Ritchie Martori Sam Roberts Miroslav Bajtos Ryan Graham
- 4. To do what? To give you a fresh pair of gloves
- 5. Let’s revisit theSTORY Let’s touch up our SOA for Mobile…and bring Glass @ work Architect Developer
- 6. The boss doeshave valid questions I recommend Node.js for our API strategy Node is nerdy like systems programming !...Where are tools ? Do we have an IDE ? Architect The Boss
- 7. Node.js APIs actuallyhave a life cycle 8 Reliable API Solutions powered by Node.js Compose Deploy / Scale Monitor Secure / Manage API Studio Controller StrongOps API Gateway Development Production
- 8. Let’s start bydebugging • Chrome debugger • Debug Remotely on Production • Breakpoints & uncaught exceptions • Source Maps • Unit test integration • Display/edit variables • Pre-loaded breaks Known as Node-Inspector / StrongLoop-Debug
- 9. Debugging server sideclustered apps Debugging single process slc debug app.js Debugging clustered process in V11 slc debug app.js http://localhost:8080/debug?port=<5858+ID-of-process> Debugging clustered process in V10 %node debug –p <PID> connecting…ok … debug>quit %node-inspector http://127.0.0.1:8080/debug?port =5858 Or process._debugPort = 5858 + cluster.worker.id
- 10. Can Errors bestitched into a trace and recovered ? Enter StrongLoop / Zones
- 11. But is itenough to convince him ? Plus I can now diagnose CPU hotspots and Memory Bottlenecks Node is fast you said !...what bottlenecks and leaks ? Architect The Boss
- 12. How to diagnoseCPU hotspots upto line of code ? Enter StrongLoop OSS - Node Profiler slc runctl cpu-start/stop PID >> CPU Proc files
- 13. Can I visuallysee my hotspots ? Chrome Dev Tools – For Server !
- 14. How about afancy chart for my boss ? Call Stack Path and function CPU Cycle times
- 15. Does Node havememory leaks ? StrongLoop OSS - HeapSnapshots slc runctl heap-snapshot PID
- 16. heapdump for V8snapshots npm install heapdump Add to app : var heapdump = require(‘heapdump’) Method 1 : writeSnapshot var heapdump = require('heapdump') ... heapdump.writeSnapshot() Method 2 : SIGUSR2 (Unix only) kill –USR2 <pid> Make sure your directory is writable process.chdir('/path/to/writeable/dir’)
- 17. Heap Snapshot strategies Programmatic heap snapshots (timer based) var heapdump = require('heapdump') ... setInterval(function () { heapdump.writeSnapshot() }, 6000 * 30) <strong>(1)</strong> Programmatic heap snapshots (threshold based) var heapdump = require('heapdump') var nextMBThreshold = 0 <strong>(1)</strong> setInterval(function () { var memMB = process.memoryUsage().rss / 1048576 <strong>(2)</strong> if (memMB > nextMBThreshold) { <strong>(3)</strong> heapdump.writeSnapshot() nextMBThreshold += 100 } }, 6000 * 2) <strong>(4)</strong>
- 18. Another fancy chartplease
- 19. We are gettingthere… See we have great Dev Tools Not at 3:00 a.m. in the night !….Show me its Production ready Architect The Boss
- 20. Ok…here is 24x7monitoring
- 21. On-Premises / 3rdParty monitoring ? 22 • Graphite • Splunk • Datadog • Introscope • Others
- 22. And On-Demand DynamicInstrumentation Monkey Patching Agent App Dynamic Instrument Agent App • Live Edit • Line Level Instrumentation • Any package, any framework, any code • Custom logic • Counters, Gauges and Timers • HA rollback
- 23. Almost there … Look I got monitoring Good…but before I promote you, show me some scale Architect The Boss
- 24. Vertical and pseudo-horizontalscaling • OSS Controller • Cluster Mgmt. • Hot Deploy • Rolling Restart • Cluster State Mgmt. • Process Mgmt. • Deploy
- 25. MESH – Allthings distributed • Distributed Deploy • Cross MC state • HA • Docker Containers • Auto Scaling • Process Mgmt.
- 26. Log Aggregation •OSS Controller • Structured Logging • Log Aggregation • 3rd Party Integration
- 27. That’s how wewin. hehe…I am a Node Hipster now ! The Boss
- 28. Nodies are notjust silicon valley hipsters ! And most recently…. #1 Retailer
- 29. JUST WIN, BABY! First there was Node Thank you!