I came home on Easter Sunday and my desktop was off.
Not sleeping. Not locked. Off. Black screen, no fans, nothing. I hadn’t shut it down. I’d been out since 7:30 AM volunteering at church. I came home to a dead machine and zero explanation.
So I did what any reasonable person does when their Linux PC mysteriously shuts down.
I investigated it like a crime scene.
Step 1: Establish a Timeline
The first thing you do at any crime scene is figure out when it happened. Linux keeps a log of every boot and shutdown in /var/log/wtmp. The last command reads it:
last -x shutdown rebootreboot system boot 6.19.9-arch1-1 Sun Apr 5 18:29 still running
reboot system boot 6.19.9-arch1-1 Wed Apr 1 11:17 still running
shutdown system down 6.19.8-arch1-1 Sun Mar 29 22:19 - 22:20 (00:01)
...Two things jumped out immediately.
The current boot started at 6:29 PM. That’s when I got home and turned it on. The previous boot started on Wednesday, April 1st at 11:17 AM — and has no end time. No shutdown entry. Just still running, hanging in the log like an open case file.
That gap — April 1st to April 5th — is the window. My PC ran for four days straight, then disappeared from the record.
(Screenshot: last output in terminal)
Step 2: Check for a Cause of Death
A missing shutdown entry means one of two things: the system crashed, or it lost power with no warning. Either way, there should be evidence in the system journal.
Linux’s journalctl stores logs from every boot separately. Boot 0 is the current one. Boot -1 is the previous one — the dead session. I jumped to the end of it:
journalctl -b -1 -e --no-pagerHere’s what the last entries looked like:
Apr 05 14:10:11 kernel: veth43a02fe (unregistering): left allmulticast mode
Apr 05 14:10:11 kernel: veth43a02fe (unregistering): left promiscuous mode
Apr 05 14:10:11 kernel: br-a881f7485680: port 1(veth43a02fe) entered disabled state
Apr 05 14:10:11 systemd-networkd: veth43a02fe: Link DOWN
Apr 05 14:10:11 systemd[1]: run-docker-netns-7d9fd7231eee.mount: Deactivated successfully.Just Docker cleaning up a container. Normal activity. Then — nothing. The journal ends at 14:10:11 and never comes back.
No shutdown target reached. No “Stopping services.” No “Power-Off.” When a system shuts down cleanly, journald logs a whole sequence — it’s verbose, orderly, almost ceremonial. This had none of that.
The journal was cut mid-sentence. The machine didn’t shut down. It was killed.
(Screenshot: journal tail in terminal — the abrupt end)
Step 3: Look for Witnesses
Time of death: somewhere between 2:10 PM and 6:29 PM on April 5th. Now I needed to figure out why.
I went back to the journal and searched for anything unusual in the hours before the cutoff:
journalctl -b -1 --no-pager -S "2026-04-05" | grep "wlan0"Apr 05 03:56:08 kernel: wlan0: Limiting TX power to 24 (24 - 0) dBm
Apr 05 03:56:41 kernel: wlan0: Limiting TX power to 24 (24 - 0) dBm
Apr 05 03:57:03 kernel: wlan0: Limiting TX power to 24 (24 - 0) dBm
Apr 05 04:34:48 kernel: wlan0: Limiting TX power to 24 (24 - 0) dBm
Apr 05 07:04:12 kernel: wlan0: Limiting TX power to 24 (24 - 0) dBm
Apr 05 10:09:48 kernel: wlan0: Limiting TX power to 30 (30 - 0) dBm
Apr 05 12:12:29 kernel: wlan0: Limiting TX power to 30 (30 - 0) dBm“Limiting TX power” is the WiFi card renegotiating with the access point. It fires when the router reboots, loses power, or reconnects. Seeing it over and over through the early morning hours means the router was cycling — coming up and going back down, repeatedly, for hours.
That’s a router on unstable power.
(Screenshot: WiFi journal entries — the pattern)
Step 4: Find the Corroborating Evidence
Here’s where the investigation got interesting. I had logs from my machine — but logs from other systems were sitting in my inbox the whole time.
My ISP, Spectrum, sent me four emails while I was at church:
- 6:00 AM — “Service Alert: We have detected an outage affecting Spectrum service in your area.”
- Updated restoration estimate: 10:00 AM
- Updated restoration estimate: 12:00 PM
- “Your Service is Restored”
I hate Spectrum. I’ve had fiber at my last few apartments and going back to cable feels like moving from a highway to a dirt road. It drops, it throttles, it sends you passive-aggressive restoration estimate emails in 2-hour increments. This outage wasn’t surprising — it was Tuesday.
But the timeline fits perfectly. Spectrum’s local node lost power in the early morning hours. My router started cycling — hence the WiFi renegotiation events from 3:56 AM onward. Service came back around noon. And then at 2:10 PM, my machine went dark.
Three independent data sources. One conclusion.
The Verdict
Cause of death: power outage.
Probably a brief one — the kind that trips a breaker, drops the whole floor, and comes back before anyone notices. My PC has no UPS (uninterruptible power supply), so it just… died. No warning, no graceful shutdown, no last words.
The Spectrum outage from the same morning wasn’t a coincidence — their local infrastructure and my apartment building were almost certainly hit by the same event.
My machine was innocent. The building did it.
What I’m Doing About It
Nothing, for now. A UPS is on the list — somewhere between “file taxes” and “cancel Audible.” But honestly, the more interesting takeaway is that between last, journalctl, and my email inbox, I had everything I needed to reconstruct exactly what happened to my machine while I was gone.
Your computer keeps meticulous records. Most people never look at them.
Next time yours goes dark without explanation — you know where to start.