Intervention commands
The sixteen intervention commands in SysAdmin Crisis: what each one changes in the fleet, which ones move money and which take time to work.
Last updated:
Interventions change the world. Most are free, five move money, and several take time to take effect. This page describes what each one does — matching them to incidents is the part of the game worth keeping.
The table
| Command | Effect | Cost |
|---|---|---|
kill <pid> | Terminate a process by PID | — |
kill_query <id> | Terminate a database query by id | — |
restart <host|service> | Restart a host or service; 10–30s of downtime | — |
block_ip <ip|cidr> | Drop traffic from an address or CIDR block at the firewall | — |
unblock_ip <ip> | Lift a firewall block | — |
enable_ratelimit | Turn rate limiting on; trims 10% of legitimate traffic too | — |
disable_ratelimit | Turn rate limiting back off | — |
scale_up <role> | Add a server to a role; ready in 30s | $100 |
scale_down <role> | Remove a server from a role | +$40 |
drain <host> | Move traffic off a host without stopping it | — |
undrain <host> | Put a drained host back in the pool | — |
clean_logs <host> | Free disk space by rotating and deleting old logs | — |
expand_disk <host> | Grow a host's disk permanently | $50 |
fix_cooling <host> | Repair the cooling on a host | $75 |
renew_cert | Renew the TLS certificate; takes 10s | $20 |
disable_cron <name> | Disable a scheduled job so it stops firing | — |
Reversible versus structural
The free commands are mostly reversible, and the paid ones are mostly permanent. That is the shape of the economy: you can experiment cheaply with things you can undo, and you pay for changes that raise a ceiling.
unblock_ip exists specifically so a false positive is recoverable —
blocking the wrong address is a mistake you can walk back. Rate limiting can be
turned off again, and a drained host can be returned to the pool. Growing a
disk or repairing cooling cannot be undone, and should not need to be.
Things that do not happen instantly
| Command | Delay |
|---|---|
scale_up | 30 seconds before the new server serves traffic |
restart | 10–30 seconds of downtime on the target |
renew_cert | 10 seconds before the certificate is live |
Costs that are not money
enable_ratelimit trims 10% of legitimate traffic along with the rest, so it
is never free even though it costs nothing. restart buys a clean state with
guaranteed downtime. drain moves load somewhere else, which means somewhere
else now carries it.
A wrong intervention costs 10 score points, and any money it spent is gone whether it helped or not. The two penalties compound quietly across a night — which is the argument for running a diagnostic first, every time.