Hi @jeremieQuiviger UF relay layer is almost certainly the cause of your queue saturation, and the architecture itself is the root problem. Heavy Forwarder output capacity is substantial — it's a full Splunk instance with configurable output queues, TCP connection pooling, and parallel pipelines. but the UF's are designed to be sources not relays in your case HF pushes data at full speed, the UF relay can't keep up, and backpressure propagates back to fill the HF's output queue. as @PickleRick said technically splunk allows HF->UF relay->Indexers but its not supported and recommended for Prodcution it should be like UF (collect) → HF (transform) → HF relay → Indexers
... View more