Real systems. Misleading symptoms. Evidence first.
Real IT Troubleshooting Stories and Sysadmin Field Notes
Real-world IT troubleshooting stories about servers, networks, security, backups, automation, and the failures that only make sense after the evidence arrives.
Latest failure · November 2026My UniFi NVR Had 8 CPU Cores. One of Them Was Doing All the Work.
Camera live view froze and stuttered for everyone. New PC, new cable, new switch port: nothing. The real problem was one CPU core inside the NVR choking on every network packet, and one Linux setting fixed it.
Read the whole mess →Failure reports worth keeping
Quiet Alerts Are an Engineering Product
A system sent hundreds of versions of the same warning and proved that notification volume is not the same thing as safety.
Open report →Why I Don't Take Phone Calls (It's Not Because I'm Difficult)
I can't hear you well, I won't remember it all, and I'll say the wrong thing on the spot. Then a client proved the fifth reason for me.
Open report →Proof, Not Green Lights
APIs, queues, processes, builds, and dashboards kept telling me yes to questions I had not actually asked.
Open report →Partial Files Are Not Finished Files
A scanner found an audio file while it was still being written and taught me that a filename is not a completion receipt.
Open report →Packaging Is the Final Boss
The game worked, the archive looked complete, and one wrong launcher path made a good build appear dead.
Open report →Last Good Data Is a Feature
When a public-data source breaks, replacing yesterday's truth with today's garbage is not freshness. It is vandalism with a timestamp.
Open report →These are not clean-room tutorials built around a predictable demo. They are stories from real technical work: the outage that looked like hardware but was software, the alert that worked perfectly and reached nobody, the backup that behaved exactly as configured and still caused damage, and the API that turned one innocent assumption into a very long night.
Every field note follows the evidence. The useful part is rarely the final command by itself. It is the misleading symptom, the tempting wrong answer, the check that changed the direction of the investigation, and the boundary that keeps the same failure from returning.