2026-06-17T12:03:04.189735Z ERROR vector::topology::builder: Configuration error. error=Sink "backflush-bucket": error occurred when building buffer: failed to build individual stage 0: failed to seek to position where writer left off: failed to validate the last written record: failed to decoded record: InvalidProtobufPayload internal_log_rate_limit=false
2026-06-17T12:03:04.189758Z ERROR vector::topology::builder: Configuration error. error=Sink "unified-firewall": error occurred when building buffer: failed to build individual stage 0: failed to seek to position where reader left off: failed to decoded record: InvalidProtobufPayload internal_log_rate_limit=false
stream closed: EOF for telemetry-hubs/vector-1 (vector)
It already happened more than twice. We're running Vector in Kubernetes and are also using disk buffering via PVCs. The only resolution for now was killing Pods with PVCs which results in data loss so thats not really an option.
2026-06-17T12:03:03.373025Z INFO vector::app: Log level is enabled. level="info"
2026-06-17T12:03:03.374957Z INFO vector::app: Loading configs. paths=["/etc/vector/vector.yaml", "/etc/vector/dynamic"]
2026-06-17T12:03:03.390993Z WARN vector::config::loading: Transform "route-config._unmatched" has no consumers
2026-06-17T12:03:03.391017Z WARN vector::config::loading: Internal log [Transform "route-config._unmatched" has no consumers] is being suppressed to avoid flooding.
2026-06-17T12:03:03.391039Z INFO vector::app: Starting watcher. paths=["/etc/vector/vector.yaml", "/etc/vector/dynamic"]
2026-06-17T12:03:03.391058Z INFO vector::app: Components to watch. paths=[ComponentConfig { config_paths: [], component_key: ComponentKey { id: "attributor_logs" }, component_type: Transform }, ComponentConfig { config_paths: [], component_key: ComponentKey { id: "backflush-transform" }, component_type: Transform }, ComponentConfig { config_paths: [], component_key: ComponentKey { id: "route-config" }, component_type: Transform }, ComponentConfig { config_paths: [], component_key: ComponentKey { id: "seahash-config" }, component_type: Transform }, ComponentConfig { config_paths: [], component_key: ComponentKey { id: "prom_exporter" }, component_type: Sink }, ComponentConfig { config_paths: [], component_key: ComponentKey { id: "backflush-bucket" }, component_type: Sink }, ComponentConfig { config_paths: [], component_key: ComponentKey { id: "shard-7f80aec0-2322-4947-a9c4-8dfd4a33dff7" }, component_type: Sink }, ComponentConfig { config_paths: ], component_key: ComponentKey { id: "unified-firewall" }, component_type: Sink }]
2026-06-17T12:03:03.391069Z INFO vector::config::watcher: Creating configuration file watcher.
2026-06-17T12:03:03.392867Z INFO vector::config::watcher: Watching configuration files.
2026-06-17T12:03:04.189735Z ERROR vector::topology::builder: Configuration error. error=Sink "backflush-bucket": error occurred when building buffer: failed to build individual stage 0: failed to seek to position where writer left off: failed to validate the last written record: failed to decoded record: InvalidProtobufPayload internal_log_rate_limit=false
2026-06-17T12:03:04.189758Z ERROR vector::topology::builder: Configuration error. error=Sink "unified-firewall": error occurred when building buffer: failed to build individual stage 0: failed to seek to position where reader left off: failed to decoded record: InvalidProtobufPayload internal_log_rate_limit=false
stream closed: EOF for telemetry-hubs/vector-1 (vector)
A note for the community
vconf.zip
Problem
We have a working configuration were we use
opentelemetryas a source and twoaws_s3as a sinks. Vector compiles it correctly but out of nowhere it starts to break with the following output:It already happened more than twice. We're running Vector in Kubernetes and are also using disk buffering via PVCs. The only resolution for now was killing Pods with PVCs which results in data loss so thats not really an option.
vconf.zip
Configuration
Version
0.55.0
Debug Output
Example Data
Additional Context
No response
References
No response