Infrastructure as code might be literally impossible
summary
Argues that reproducible infrastructure fails because core tools break in concrete ways: hardware atomics on an Opteron revision fail so MySQL detects and warns; the Ruby VM sometimes frees live objects causing Chef/Puppet to segfault and the practical workaround is to disable the garbage collector; package managers yum/apt routinely miss or mishandle GPG verification (pyGPGme dependency, repo vs package checks, GPGv3 quirks) and mirrors can be poisoned or cause disk‑filling DoS; create‑repo metadata parsing corrupts versions on CentOS 7.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:25What is code
- 8:49Infrastructure code
- 10:48Puppet bug
- 15:14yum and GPG
- 22:02AWS CA revoked
- 26:17Fake mirror attack
- 28:25Incorrect package metadata
- 31:37Conclusion
- 33:58Questions
- 34:15How to define a better system
- 36:08Should we keep using Linux
- 38:07Is it impossible
- 39:08Python
- 40:28Unit tests
- 42:07Research university
- 42:46Docker
related talks
41mInfrastructure As Code Might Be Literally Impossible
54mInfrastructure as code (anti) Patterns
36mInfrastructure as Code Done Right: Discovering and Fixing Common Missteps
46mImplementing Infrastructure as Code
05mProgramming your Infrastructure
28mFrom Monitoring to Automated Testing of Your Infrastructure Code
46mThink of your infrastructure as code
46mThe Rise and (Potential) Fall of Infrastructure as Code