Airbyte can't sync any data. but no error is thrown. It says 0 bytes ad no records loaded after 30 minutes. I have to restart airbyte-workload-launcher to fix it. but why it is happening?

Based on the Airbyte documentation and community issue reports, this problem typically occurs when the Workload Launcher fails to properly launch or track the execution pods, or loses communication with Temporal / the database, causing jobs to get orphaned and eventually close out without committing records [Github Issue (open): sync pulling 0 records & exits out without raising error; Github Issue (open): sync pulling 0 records & exits out without raising error]. Here are the most common reasons why this happens: 1. Workload Lifecycle / Timeout Expiration Airbyte uses a Workload Monitor background process that tracks whether workloads are making expected progress. If an attempt dies early (for example, due to a pod eviction, OOM kill, or container exit) or gets stuck before properly reporting state, Airbyte may fail to schedule retries and end up marking the job as completed with

0 bytes | no records loaded

[Github Issue (open): Sync job marked succeeded with 0 records loaded when its only attempt dies early (no retry scheduled); Documentation Platform > Understanding-airbyte > Workloads & jobs > Workload launching architecture>]. 2. Temporal Persistence Timeouts & State Loss When the Temporal orchestrator experiences database stress (e.g., PostgreSQL persistence timeouts such as

persistence.AppendHistoryTimeoutError

or

context deadline exceeded

), it cannot persist workflow state updates [Github Issue (open): sync pulling 0 records & exits out without raising error]. • By the time the Launcher attempts to update the status to

LAUNCHED

, the workload record might have timed out or been cleaned up (returning

io.airbyte.api.client.ApiException: Gone

) [Github Issue (open): sync pulling 0 records & exits out without raising error; Github Issue (open): sync pulling 0 records & exits out without raising error]. • Restarting the

airbyte-workload-launcher

temporarily clears out stuck in-memory states and resets queue consumers. 3. Resource Bottlenecks on Kubernetes If your deployment experiences resource pressure, the launcher may fail to start the replication pod (or the pod may remain in

Pending

/ get dropped without raising an explicit failure), throwing errors such as

ResourceConstraintException: Unable to start the REPLICATION pod

[Github Issue (open): sync pulling 0 records & exits out without raising error]. • If containers take too long to launch (due to image pulling or waiting for Kubernetes autoscaling), the workload may timeout before replication starts [Documentation Platform > Understanding-airbyte > Workloads & jobs>]. 4. Dataplane Configuration Mismatches In some Airbyte versions, workspaces pointing to a missing or mismatched dataplane group (e.g., after upgrading without configuring

global.dataplaneGroups.defaultDataplaneGroupName

) cause jobs to start the discovery step but fail to claim and launch orchestrator, source, or destination pods, resulting in silent

0 bytes

completions [Github Issue (open): airbyte services stuck where no new job pods are launched]. How to Diagnose and Prevent It 1. Check Workload Launcher Logs: Look for

ResourceConstraintException

,

ApiException: Gone

, or

Failed to update workload status to launched

[Github Issue (open): sync pulling 0 records & exits out without raising error; Github Issue (open): sync pulling 0 records & exits out without raising error]. 2. Inspect Temporal and Database Health: Check the CPU, memory, connection limits, and disk I/O on the database backing Temporal and Airbyte [Github Issue (open): sync pulling 0 records & exits out without raising error]. 3. Inspect Pod Scheduling: Run

kubectl get pods

and

kubectl describe pod <replication-pod-name>

to see if replication pods are failing to schedule due to CPU/memory limits [Documentation Platform > Understanding-airbyte > Workloads & jobs > Workload launching architecture>]. 4. Tune Concurrency Limits: Lower concurrency via

WORKLOAD_LAUNCHER_PARALLELISM

or

MAX_SYNC_WORKERS

to prevent overloading the workload launcher and Temporal cluster [Github Issue (open): sync pulling 0 records & exits out without raising error; Documentation Platform > Understanding-airbyte > 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.