Stream compaction policies: bounded history without touching the read side New in 1.14.0
Declare and operate CritterWatch stream compaction policies — cron-scheduled, threshold-driven background compaction across a fleet — and read what they have actually done.
Use when event streams have grown unbounded, when deciding whether compaction is safe for an aggregate, when an operator asks why a policy has not run, or when an agent must run, pause or resume one over MCP. Covers the policy model (aggregate type + growth threshold + cron), why the threshold measures growth since the last compaction rather than raw version, the dry-run-until-armed safety rule and why arming is deliberately a code change, MaxStreamsPerTick, tenant scope, the timeline event types (CompactionDryRun / CompactionRun / CompactionRunFailed), the five MCP tools and their honesty envelopes, the store support matrix (Marten and Fisher compact; Polecat selects correctly and then refuses), and why a policy may name an aggregate the application does not snapshot.
Skill details
- Skill name:
wolverine-integrations-critterwatch-stream-compaction - Applies to:
wolverine,marten,polecat,fisher - Tags:
observability,critterwatch,mcp,event-sourcing,event-store,compaction,archiving,scheduling,cron,retention,agents - Added in: 1.14.0
- License: Proprietary. Copyright JasperFx Software LLC.