8 timed incidents on real virtual machines. Each one boots genuinely broken: a service is down, and the cause is somewhere underneath it. Nothing is simulated, and there is no multiple choice.
A service fails because a mount failed because a disk lied. Restarting the service was never going to work, so you walk the chain down. The map only shows what you have proven: run a real diagnostic and the facts light up one node at a time.
Three hints exist per incident and each one costs you. Every solved run records a time, and there is one leaderboard per incident, so your best stands until somebody takes it.
These are system problems rather than algorithmic puzzles: log parsing, mounts, systemd units, signals, and the places Linux hides its evidence.
vault.service is down and customer records are unreachable. Find the root cause and bring the vault back online. Hints cost time. Timer running.
app.service is down and the order archive under /srv/app is gone. The storage stack was hand-built: a disk image, a loop device, LVM on top. Somewhere in that chain a link is missing.
reportd.service refuses to start and the quarterly archive is unreachable. More than one thing is broken, and fixing the first fault only gets you to the second.
mailq-drain.service is down on a mail relay and nothing can write to the spool, while the volume reports 905M free and 1% used. Three customer messages are stuck in the queue. Hints cost time. Timer running.
The nightly backup works every time it is run by hand and has never once worked on schedule. Cron fires it every minute, nothing errors anywhere visible, /srv/backups stays empty, and empty archives pile up at the filesystem root. Hints cost time. Timer running.
paymentsd is down since a hardening change and cannot read a config file it owns. The file is unchanged: mode 644, right owner, and root reads it without error. The service account can see the file exists and still cannot open it. Hints cost time. Timer running.
First night on the floor. The vault server is locked, the day shift is sloppy, and the password is written down somewhere on ws-06. Look around, find the key, let yourself in.
The vault key fits the deep-storage door. The archive under the floor is a maze of rows and lockers, the lights are bad, and Reyes kept a locker down there. pwd is the wall map. cd is your feet.
Playing an incident needs an account. Start with the training tracks if you want the underlying skills first.