Videos / DevOps

Episode 3 of 14 · DevOps, end to end

Process management and reverse proxies: keeping it running, and in front of it

Chapters

1. The two jobs nobody wrote0:00
2. Fork, exec, and a parent0:15
3. The tree, and PID one0:33
4. Orphans go to the root0:50
5. Zombies, and who reaps them1:07
6. What an init system is for1:23
7. A unit file, in full1:41
8. The service section, line by line1:58
9. Restart, and what counts as failure2:16
10. The loop that never gives up2:34
11. Enable is not start2:52
12. Reading the journal3:11
13. Alive is not the same as ready3:28
14. Liveness: restart me3:44
15. Readiness: skip me for now4:01
16. The probe that took everything down4:21
17. Slow starts and the kill loop4:40
18. Shutting down on purpose4:58
19. The desk in front of the building5:17
20. Three ports, one door5:34
21. How the routing actually works5:54
22. The address the app now sees6:11
23. One name, many backends6:30
24. The door is now the risk6:47
25. One place that holds the certificate7:04
26. Routing by host and path7:22
27. Buffering the slow client7:40
28. Compress and cache at the edge7:58
29. Rate limiting, per key8:17
30. Shedding load on purpose8:35
31. Every call needs a deadline8:54
32. The timeout budget9:10
33. Keep-alive, and the mismatch9:26
34. The retry that doubles the load9:43
35. Twenty-seven requests from one10:00
36. Budgets, backoff, and a breaker10:19
37. Keeping it running10:37

Process management and reverse proxies: keeping it running, and in front of it. Two things stand between your code and a working service. Something that keeps the process running when it dies. And something that stands in front of it, so the outside world only

Mark as watchedSaved in this browser only.