Honeypot Live Stats — 82 connections today from 31 unique IPs (1096 total, 73 unique IPs all-time)
Miles Huynh
· Reading time: 3 minutes · Admin

Docker, Index, and a very expensive lesson in reading pricing tables

Quiet learning day, mostly — until AWS almost let me subscribe to something that costs $57,600 a month.

Dockerizing Bludit, locally

Containerized nginx + PHP-FPM + a fresh Bludit clone with Docker Compose — but on my own machine, not the live EC2, after last week's MySQL RAM incident taught me not to experiment on the real box. Hit one snag (missing PHP gd extension, fixed with a small custom Dockerfile), then a useful realization: Docker doesn't save disk space. Each container carries its own copies of libraries, so my current bare-metal setup is actually leaner. Docker's value is isolation and reproducibility, not size.

What an Index actually buys you

Built 300,000 fake rows in a local MySQL container to see indexing in action. Query time looked identical before and after (0.00 sec both times) — but EXPLAIN told the real story: type: ALL (full scan) became type: ref (index lookup). The gap is invisible at this scale and enormous at millions of rows. Lesson: judge an index by the query plan, not the stopwatch.

The $80-an-hour near-miss

Tried wiring up AWS Bedrock to help draft posts. Hit a token quota error, went looking for a fix, and landed on a "Purchase options" page that looked like routine setup. It wasn't — it was Provisioned Throughput pricing, $80–99 per hour, not the per-token rate I'd already checked. One click from a five-figure monthly line item. Backed out, no charge.

Turns out I didn't need Bedrock at all — I already have Claude through my existing subscription, right here in this chat. That's the whole solution for "help me write" or "summarize this log." Bedrock only earns its complexity when something needs to run unattended on a schedule.

What I actually learned

  • Docker isolates; it doesn't shrink disk usage.
  • Check EXPLAIN, not the clock, when judging an index.
  • On AWS Marketplace, two very different products can wear the same name — read every number before clicking Subscribe.
  • The simplest tool is often the one already open in front of you.