Controlled infrastructure experiments. Load testing. Bottlenecks. Failure analysis. Linux • DevOps • Performance piped.video/@SystemAxis
- Tweets 279
- Following 278
- Followers 29
- Likes 594
A Linux service can be “active (running)” and still be completely unreachable.
Before digging through logs, check the listener:
ss -lntp
I’ve seen the service configured for 8081 while everything else expected 8080.
Running process ≠ working service.
Made with AI
1
22
Docker didn't make applications easier to run.
It made environments predictable.
The real problem was never the code.
It was "works on my machine."
1
7
The more Linux I learn, the less I care about shiny features.
Stability, visibility, and knowing why something broke are worth a lot more.
1
16
One of the most Linux things ever: Everything looks fine until you hit the file descriptor limit and suddenly half the system starts failing in weird ways.
ulimit -n has humbled many engineers.
1
25
The more time I spend with Linux, the more I understand why engineers love it.
Not because it’s “cool.”
Because it gives you control, visibility, and fewer layers between you and the system.
2
21
Linux teaches you how systems actually work. DevOps teaches you why they fail.
Both together change how you think about infrastructure.
1
32
This is where a $6 server started collapsing.
1 vCPU / 1GB RAM
Nginx → Gunicorn → Python
k6 load test
~200 users → 1700 req/s
~1000 users → ~500 req/s
CPU 100%
4096 TIME_WAIT
connection resets
The hardware wasn’t the problem.
Full breakdown: piped.video/watch?v=EtHRR_GU…
1
36
DevOps isn’t just CI/CD and YAML files.
It’s ownership, reliability, and fixing things when production breaks at 3am.
1
23
DevOps thought:
If you want to understand systems,
run Linux.
If you want to understand reliability,
run it in production.
1
29
Linux thought:
The more you learn how the system works,
the more powerful it becomes.
Linux rewards curiosity.
2
32
Linux doesn’t try to hide how the system works.
It lets you understand it.
That’s why engineers keep coming back to it.
1
31
If AI becomes a metered utility, reliability, latency, and cost control become DevOps problems.
Someone still has to run the infrastructure that delivers every ‘thought’.
1
31
Engineering wisdom:
Tools change every few years.
Understanding systems, failure modes, and real user behavior stays valuable for decades.
1
25
DevOps reality:
Most outages start as something small.
A slow query.
A retry loop.
A queue that quietly fills up.
Then everything fails at once.
1
27
DevOps reality:
The system usually doesn’t fail during deployment.
It fails three hours later,
when real traffic hits a code path nobody tested.
1
20
Most infrastructure failures are not caused by scale.
They happen because nobody tested the system under real load.
Production is the worst place to discover your bottleneck.
1
18
Many DevOps problems are not tool problems.
They are system understanding problems.
If you don’t know where the bottleneck is,
more Kubernetes, more cloud, or more hardware won’t fix it.
1
27