End of story.
Empty. The core database had been deleted. Not corrupted. Not unmounted. Deleted. And the last access timestamp on the parent directory was 2:46:58 AM—one second before the first alert.
$ ls -la /dev/vault/ total 0 drwxr-xr-x 2 root root 60 Jan 17 2022 . drwxr-xr-x 22 root root 4096 Jan 17 2022 ..
Except for last Tuesday, when he'd left it in his desk drawer for two hours during the all-hands meeting.
"We rebuild. We tell them it was a hardware failure. RR-4036. Database connection error. Force majeure. We restore from the transaction logs—the ones I have on a private drive."
But when he plugged in his diagnostic monitor, the screen showed only:
"Why?"
"So we lie?"
Marcus looked at his phone. Three missed calls from her. The fourth was ringing.
"Marcus, I need you to wipe the forensic logs. All of them." Her voice was calm. Too calm.
He logged in remotely.
Outside, a floorboard creaked.
The terminal blinked back. $ systemctl restart edtm-db-listener Failed: Unit edtm-db-listener.service not found. He frowned. Not found? That was impossible. The listener was a core daemon. He checked the process list. Nothing. He checked the database directory. Also nothing.
By 3:15 AM, Marcus was in the data center, the cold air raising goosebumps on his arms. The primary database server—a hulking Dell PowerEdge—was still running. Its fans whirred. Its lights blinked green.
Not a crash. An absence.