I’ve been working with UNIX/Linux for 30–40 years.
You’d think by now I’d have seen every trap.
Turns out — some of the oldest ones are still the sharpest.
The Setup
I’ve got a simple media pipeline:
/export/media→ working path- backed by a 4TB disk mounted underneath
- feeding into a broader pipeline (sync, publish, etc.)
Except… the disk wasn’t mounted when I first built it.
Linux didn’t complain.
It just quietly wrote everything to the root filesystem instead.
No warnings. No errors. Just:
“Sure, that path exists — I’ll use it.”
The Moment
You spot it when something feels off.
Disk usage doesn’t make sense.
Paths look right… but something’s wrong.
Then it clicks:
“That wasn’t mounted when I started…”
Yeah. That one.
The Fix (Old Trick, Still Gold)
Copy it properly:
cd /export/media && \
tar cf - . | (cd /disk/EXT_4TB_A/MEDIA/media && tar xpf -)
Clean. Streaming. No temp files. Muscle memory.
The Lesson (This Is the Important Bit)
This wasn’t really a mistake.
It was a missing guardrail.
Linux does exactly what you tell it — even when that’s not what you meant.
So the real fix isn’t the copy.
It’s making sure this can’t happen again.
One Line That Changes Everything
chmod 000 /export/media
Now:
- If the disk is mounted → everything works normally
- If it’s not → writes fail immediately
No silent fallback. No guessing. No surprises.
Why It Matters
This isn’t about old hardware.
It’s not about flaky USB drives or repurposed machines.
A brand-new NAS would behave the same way.
The difference isn’t the gear.
It’s whether your system:
fails silently
or
fails loudly and correctly
Still Learning
After decades in this space, you don’t stop learning.
You just stop learning new tricks —
and start rediscovering the old ones that still matter.
This old dog is still learning old tricks.