GorillaKill

The "gros bourrin" scenario: no survivors, no PDB negotiation, just a mass Delete of every pod matching the configured selectors.

When to use it

• You want to simulate a full-zone outage or a simultaneous loss of every replica of a workload.
• You want to validate cold-start recovery, leader re-election, or cache warm-up paths.
• You explicitly do not want PDB protection.

For a steady drip of single-pod kills with a safety floor, use Killing instead.

Minimal config

Run once at startup, kill everything, then idle:

kind: GorillaKill
metadata:
  name: gorilla-my-app
  namespace: default
scenario:
  when: once
  matchers:
    labels:
      app: my-app

Or on an interval:

kind: GorillaKill
metadata:
  name: periodic-zone-loss
  namespace: default
scenario:
  when: periodic
  interval: 10m
  matchers:
    labels:
      app: my-app

Or on a cron schedule:

kind: GorillaKill
metadata:
  name: nightly-zone-loss
  namespace: default
scenario:
  when: periodic
  cron: "0 3 * * *"    # every night at 03:00
  matchers:
    labels:
      app: my-app

Full reference

kind: GorillaKill
metadata:
  name: gorilla-my-app                  # required
  namespace: default                    # required
scenario:
  when: once                            # required — "once" | "periodic"
  interval: 10m                         # required when when=periodic and no cron
  # cron: "0 3 * * *"                  # alternative to interval — standard 5-field cron
  wait: 30s                             # optional — delay before first/only run (ignored with cron)
  dryRun: false                         # optional — log only
  matchers:                             # required — at least one selector
    labels:
      app: my-app

Field details

Field	Type	Default	Notes
scenario.when	enum	—	Required. once runs a single time then exits the module loop. periodic ticks on interval or cron.
scenario.interval	duration	—	Required when when: periodic (unless cron is set). Mutually exclusive with cron.
scenario.cron	string	—	Standard 5-field cron expression. Only valid with when: periodic. Mutually exclusive with interval.
scenario.wait	duration	0	Delay before the first (or only) run. For periodic, must be < interval. Ignored with cron.
scenario.dryRun	bool	false	When true, logs each intended kill without calling the API.
scenario.matchers	object	—	See Matchers.

Behavior

1. Collect the union of pods matching every configured selector.
2. If no pods match, emit a gorillakill skipped: no_match warning and exit.
3. For each matched pod:
   • Log the intended kill (always).
   • If not dryRun, call CoreV1().Pods(ns).Delete(...) — no eviction, no PDB.
4. Aggregate per-pod errors using errors.Join and return them so the orchestrator logs a single run failure.

Scheduling semantics

• when: once returns a ScheduleOnce to the orchestrator. The module loop runs exactly one tick, then the goroutine exits.
• when: periodic returns a SchedulePeriodic, same semantics as Killing.

Combined with wait, this lets you express "five minutes after the agent starts, kill every pod of service X" as a one-shot scenario.

RBAC

Verb
pods → list, get, delete

Plus read access for the selectors you use (Matchers RBAC).