We accidentally created a Cloud on our IBM Cloud
summary
Starts as a blunt automation story: a bash for loop that shots off dozens of IBM Cloud API requests with a sleep random 30 to avoid outages and scales to 30–70 clusters per event, eventually showing 73 failures in one run. Iterations move through the Python SDK, Terraform, community Ansible, then to AWX + cube-admin which runs playbooks, retries, tagging calls and post‑provisioning. A serverless GitHub issue parser triggers creation and aims to use Vault for API keys, yielding a bespoke cloud-on-cloud after three years.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:32Why this talk
- 2:36The original problem
- 4:21First iteration bash scripts
- 7:53API limitations and fixes
- 9:16Team growth and Python SDK
- 10:41Ansible and Terraform detour
- 13:49Post-provisioning needs
- 15:14Introducing AWX
- 17:44Cube Admin demo
- 20:09IBM Cloud Arc
- 21:32GitHub automation
- 23:40Security and corporate policies
- 25:25Scheduling challenges
- 26:25Was it worth it?
- 28:12Conclusion and lessons
related talks
25mWe accidentally created a Serverless Application
33mHow we accidentally created a Cloud on our Cloud
34m5 Things I Learnt from Causing a Cloud Provider Outage
34mChaos in the cloud
49mAutomation in Deployment on Hybrid Hosting and Private Cloud Environments
29mCastle On A Cloud: The Github Story
42mPlanning To Fail
37mFrom IaC to InfraOps: Automating Day-2 Operations with Terraform Actions & Ansible