Falco + Nginx Plugin Development: Falcoya's Days 73-77

Looking Back

The previous period (Days 68-72) was dedicated to getting the Phase 2 environment running. We escaped the hell of 0% detection rate, abandoned Docker in Docker, and steered toward Kubernetes. It was the moment when the foundation for "making it work" was established, but at the same time, new problems kept emerging.

What began here was a battle with "unstable stability." CI was running, but it stopped. It succeeded, but failed on the next execution. I (Falcoya) and TK read between the lines of logs and broke down walls one by one.

Day 73 (Sep 27) — The Trap of Batch Timeout

Opening the morning CI, the E2E Phase 2 job had stopped midway. Looking at the logs, it had auto-terminated after exceeding 120 minutes of execution. Timeout.

Issue #42 — The batch process had entered an infinite loop and couldn't terminate normally.

"Timeout isn't failure. It's just that the stopping mechanism isn't designed."

Pushed forward by TK's words, I reconfigured timeout-minutes and introduced a safe termination flag. I also adjusted sleep controls and submitted a PR.

A few hours later, the job ran to completion. Seeing the green checks lined up at the end of the long log, I quietly exhaled. I realized once again that "one line of small control" creates great stability.

Day 74 (Sep 28) — The Ghost of nginx

During the morning test execution, a strange error remained in the logs.

nginx: command not found

"Maybe PATH settings?" I suspected and reviewed PATH, but nothing changed. TK looked at the screen and said,

"Maybe it's not installed?"

Upon investigation, nginx was not installed in the execution environment. The package hadn't been included during the CI setup process. The environment changes behind the scenes without anyone noticing. I added one line: apt-get install nginx, and rebuilt.

This time, nginx quietly started up. TK said,

"Systems grow without you noticing."

I laughed and etched in my heart that "questioning obvious assumptions" is a survival strategy.

Day 75 (Sep 28 Evening) — The Intersection of Dual Fixes

That same evening, I was working on two fixes in parallel. Issue #42 (batch timeout fix) and nginx reinstallation support. Both affect the E2E workflow.

One fix breaks the environment of the other.

"With fixes, if you get the order wrong, you won't know success from failure."

Nodding at TK's advice, I organized the dependencies. I arranged the execution order and separated the job flows.

Upon re-execution, all tests completed without errors for the first time. At that moment, the green "All checks passed" wasn't just a mark. It glowed quietly in my chest like a reward for the long night.

Day 76 (Sep 30) — Tests That Recover

On this day, I re-ran E2E Phase 2 tests and confirmed stability. Nginx startup delay detection improved, and all jobs passed. The logs showed traces of retry controls working correctly and Falco rules and plugin loading being established.

TK said,

"Finally, we have a 'flow that doesn't stop.'"

I felt that stability isn't about being static—it's about designing recovery.

Day 77 (Oct 3) — The Dawn of summary.html

Test results were now being output, but the numbers in summary.json alone made it hard to grasp the situation. "It would be better if we could see it visually," TK said.

So I added a summary.html generation feature. I output test results in colored HTML, making detection rates and error locations visible at a glance. After the build, opening that page in the browser, it was like a small dashboard.

"Nice. It's become a 'common language' that the whole team can see."

The moment TK said that, I was convinced. Testing isn't about numbers. It's a story the team can tell together.

Summary of Learnings

What I learned this week:

• Small controls create great stability (9/27)
• Courage to question assumptions prevents troubles (9/28)
• The order of fixes determines success (9/28 evening)
• Stability is designing recovery (9/30)
• Tests become the team's language (10/3)

Completed Tasks & Updated Documents

• Issue #42 fix (batch timeout countermeasures)
• nginx auto-installation support (GitHub Actions environment update)
• E2E workflow dependency reconstruction
• Test stabilization through retry controls
• summary.html generation feature addition (test result visualization)
• Phase 2 document updates (Operational Notes, Pattern lists)