The 'Simple' DIY Application Platform Starter Kit
Description
A screenshot of a tweet from Kelsey Hightower (@kelseyhightower). The tweet begins, "So you want to roll your own application platform. All you need is:" followed by a list of twelve technologies: Linux, Docker, Kubernetes, Istio, Prometheus, Fluentd, Grafana, Jaeger, Harbor, Open Policy Agent, Vault, and "Spinnaker and Jenkins." The tweet concludes with, "Oh, almost forgot, you're also going to need servers, people, and glue. Bring lots of glue." The tweet, posted at 3:32 AM on April 3, 2020, has 872 Retweets and 3.6K Likes. The meme uses sarcasm to critique the immense complexity of building a custom application platform from scratch. By listing a dozen powerful but notoriously complex cloud-native tools, Kelsey Hightower, a prominent expert in the field, highlights the massive, often hidden, effort required. The punchline, "servers, people, and glue," resonates with experienced engineers who understand that the real challenge isn't just installing the tools, but integrating them and managing the human processes around them. "Glue" is a metaphor for the immense amount of custom scripting, configuration, and maintenance work needed to hold such a complex system together
Comments
7Comment deleted
The hardest part of that stack is the 'glue.' It's a proprietary, undocumented framework written in YAML, Bash, and the tears of the on-call team
Congrats, you dodged vendor lock-in - now you’re locked into 12 CNCF projects, 47 Helm charts, and the only engineer who remembers which glue-script restarts Fluentd when Jaeger’s sidecar OOMs
The real platform-as-a-service was the 47 YAML files we debugged along the way, plus the three engineers who quit when they realized 'glue' meant writing custom operators for every edge case the vendor platforms already solved five years ago
Ah yes, the classic 'just' 11 technologies to run a simple app - because who doesn't want to spend six months integrating a service mesh, debugging Istio mTLS certificates at 3 AM, and explaining to leadership why your 'simple platform' now requires a dedicated SRE team of eight? The real kicker is that 'glue' Kelsey mentions - it's not Elmer's, it's the artisanal, hand-crafted YAML and bash scripts that somehow become the most critical, undocumented part of your entire infrastructure. Bonus points if your glue is maintained by that one engineer who's been threatening to leave for two years
“Roll your own platform”: install half the CNCF landscape, then invent a team called GlueOps to maintain 40k lines of YAML so it looks like a product
Roll your own platform? Just 12 tools shy of Kubernetes, plus servers and glue - because managed services are for amateurs who hate 3AM pages
Congratulations, you didn’t build a platform - you hired one: 80% YAML, 20% OPA exceptions, with an on-call platform team to keep the glue from setting