@kapa.ai In version 1.8.5 running on k8s, how do I configure platform concurrency so not more than N concurrent syncs happen?

Thinking completed The knowledge sources do not contain specific documentation or release notes for version 1.8.5, but they do outline how Airbyte's Workloads architecture and Helm charts handle sync concurrency on Kubernetes. In the Workloads architecture on Kubernetes, platform concurrency and job launching are controlled through worker and launcher configurations: 1. Configure Maximum Concurrent Sync Jobs (MAX_SYNC_WORKERS) Airbyte controls the maximum number of concurrent sync jobs supported per worker container using the

MAX_SYNC_WORKERS

setting [Workloads & jobs; Configuring Airbyte]: • Environment Variable:

MAX_SYNC_WORKERS

(defaults to 10) [Configuring Airbyte]. • Helm values: ◦ In Helm Chart V2, set

worker.maxSyncWorkers: <N>

[Upgrade to Helm chart V2 (Core)]. ◦ In Helm Chart V1, this was passed via

worker.extraEnvs

[Upgrade to Helm chart V2 (Core)]. > Note on worker replicas: Because

MAX_SYNC_WORKERS

defines the limit per Airbyte Worker container/pod, the total concurrent sync capacity across the deployment depends on the number of worker replicas (

replicaCount * MAX_SYNC_WORKERS

) [Scaling airbyte on k8s - increased job parallelism; How to configure to have a larger number of concurrency?]. If you want a hard cluster-wide ceiling of $N$, ensure the sum of

MAX_SYNC_WORKERS

across all worker replicas does not exceed $N$ (or run 1 worker replica set to $N$). 2. Configure Simultaneous Job Starts (WORKLOAD_LAUNCHER_PARALLELISM) If you also need to restrict how many jobs the Workload Launcher can initiate simultaneously to avoid overwhelming Kubernetes with pod creation, you can configure: • Environment Variable:

WORKLOAD_LAUNCHER_PARALLELISM

(defaults to 10) [Configuring Airbyte]. • Helm values: ◦ In Helm Chart V2, set

workloadLauncher.parallelism: <number>

[Upgrade to Helm chart V2 (Core)]. ◦ In Helm Chart V1, this was passed via

workload-launcher.extraEnvs

[Upgrade to Helm chart V2 (Core)]. Under the Workloads architecture, jobs exceeding the running/starting capacity remain queued in the Workload API Server until resources become available [Workloads & jobs]. 👍 0 button 👎 0 button ℹ️ Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.