Skip to content

Cratis Chronicle configuration best practices

For repeatable Chronicle Kernel deployments, pin the server image version, mount shared configuration read-only, and supply credentials at deployment time. Keep environment-specific values outside the shared file so a storage connection or secret does not travel with it into another environment.

Use chronicle.json for shared settings and environment variables for overrides. A matching environment variable wins over the file; Configuration Precedence explains how the Kernel combines these sources.

{
"observers": {
"subscriberTimeout": 5,
"maxRetryAttempts": 10
},
"events": {
"queues": 8
}
}
Property Description
observers.subscriberTimeout Allow enough time for subscriber work at the latency of your infrastructure
observers.maxRetryAttempts Set a finite retry limit, or use 0 for unlimited retries
events.queues Spread the routing of appended events when many observers share an event sequence; see How observers use the queues

The example is a starting point, not a workload recommendation. Choose the subscriber timeout and retry limit according to how long your subscribers need to complete their work and how you intend to recover from failures. See Observers before changing these values.

Increasing events.queues spreads live appended-event routing when observers share an event sequence. It does not control replay or catch-up, which read from the event log. Use Events to decide whether changing the queue count addresses your workload.

Use a specific image version rather than latest, and mount chronicle.json read-only. Keep the image version and required environment variables with the deployment manifests so another operator can identify what is running.

Supply environment-specific connection strings through your platform’s secret mechanism rather than committing credentials to JSON. Docker Configuration shows the file mount and environment-variable forms; Production hosting covers the certificates and other requirements for starting the production image.

If schema provisioning runs separately from the Kernel, follow SQL schema operations or MongoDB index operations before reducing runtime permissions. Each guide explains what verification checks and which operations can still create schema objects at runtime.