Configuring Progressive Service Deployment for the connectware-agent Helm Chart

After a restart, Connectware agents typically try to fetch their work immediately. In large installations, many agents doing this at once can trigger a “thundering herd” effect and create severe load spikes.

Progressive service deployment allows you to stagger those fetch attempts by adding per-agent delays. Enabling it helps smooth out peaks when many agents restart simultaneously.

We recommend turning this on once your deployment reaches around 100,000 resources, where the stability gains generally outweigh the added startup delay. This is a tradeoff in stability vs. faster startup.

• To enable progressive service deployment, set protocolMapperAgentDefaults.progressiveServiceDeployment.enabled to true.

In most cases the default values are sufficient and enough to enable progressive service deployment.

The following Helm values are available in both, protocolMapperAgentDefaults and protocolMapperAgents[*]:

Helm Value	Description
progressiveServiceDeployment.enabled	Enables or disables the progressive service deployment warmup mechanism. When enabled, service deployment is gradually ramped up to avoid sudden load.
progressiveServiceDeployment.warmupDuration	Defines the duration (in minutes) for which the warmup mechanism is active. After this time expires, the delay falls back to its default ("0").
progressiveServiceDeployment.warmupMinDelay	The minimum (in milliseconds) for the delay an agent can apply during the warmup period.
progressiveServiceDeployment.warmupMaxDelay	The maximum (in milliseconds) for the delay an agent can apply during the warmup period.

Example

protocolMapperAgentDefaults:
    progressiveServiceDeployment:
        enabled: true
        warmupDuration: 5 # 5 minutes
        warmupMinDelay: 2000 # 2000 milliseconds
        warmupMaxDelay: 10000 # 10000 milliseconds