FSDP reduce_scatter backward hook timed out waiting for gradient sync

An FSDP gradient reduce-scatter did not complete within the collective timeout. Like any collective timeout this names the ranks that were WAITING, not the one that failed to arrive, one rank left the collective early or died, and the rest report the timeout. This entry explains how to confirm the cause, apply the fix, and separate it from adjacent pytorch-fsdp-ddp failures.

The literal signature is "FSDP reduce_scatter backward hook timed out waiting for gradient sync". It is a distributed training failure associated with PyTorch Distributed, FSDP, DDP, and TorchElastic. The line identifies the failing operation or subsystem, while the surrounding evidence decides whether it is the initiating fault or a downstream symptom.

An FSDP gradient reduce-scatter did not complete within the collective timeout. Like any collective timeout this names the ranks that were WAITING, not the one that failed to arrive, one rank left the collective early or died, and the rest report the timeout. The failure becomes visible at this call site because the operation first requires the missing resource, valid state, healthy peer, or correct result. Earlier log lines and a known-good control carry more causal value than the final wrapper exception.

• The workload stops or loses forward progress after emitting "FSDP reduce_scatter backward hook timed out waiting for gradient sync".
• A retry on the same configuration reproduces the failure because the causal state has not changed.
• The outer framework exception can hide the rank, node, allocation, or dependency that failed first.
• Increasing timeouts or reducing workload size can suppress the symptom without correcting the cause.

• PyTorch Distributed, FSDP, DDP, and TorchElastic
• production-shaped multi-accelerator workloads
• containerized and bare-metal deployments of the same stack

Apply the following checklist to a small reproduction: each box below is a positive signal that you are looking at this exact failure rather than a sibling in the same taxonomy.

• ✓Find the first occurrence of "FSDP reduce_scatter backward hook timed out waiting for gradient sync" and preserve at least 100 lines before it.
• ✓Identify which rank, node, device, or process emitted the first related warning.
• ✓enable the flight recorder (TORCH_NCCL_TRACE_BUFFER_SIZE=2000) and dump on timeout; it names the collective each rank was in. Confirm all ranks execute the same module order, a conditional branch that skips a layer on some ranks desynchronises the collective schedule and produces exactly this.
• ✓Repeat the same input after the targeted change and require the signature to disappear.
• ✓Resume from latest checkpoint only after the control passes.

FSDP reduce_scatter backward hook timed out waiting for gradient sync

Timestamps and exact values vary across runs, but the pattern. An info-level start, an early WARN, an ERROR carrying the symptom. Is the actual fingerprint you should alert on. The Denpex platform flags this combination automatically.

find the rank that is missing rather than debugging the ones that reported. Scan every rank's log for the earliest distinct error, an OOM, an assertion, or a segfault that predates the timeout. This changes the condition in the causal diagnosis instead of hiding the outer exception. The repeated control proves ownership before recovery from latest checkpoint.

# Preserve evidence before restarting
rg -n -i 'error|exception|timeout|failed' <log-file>
nvidia-smi
python -m torch.utils.collect_env

# Find the exact signature in the complete log
rg -n -F -- "FSDP reduce_scatter backward hook timed out waiting for gradient sync" <log-file>

Adapt the snippet to your framework. The same pattern holds for PyTorch Lightning, Hugging Face Trainer, DeepSpeed, Megatron-LM, and vLLM training wrappers. Where the wrapper exposes a config flag (for examplelr_scheduler_type in Trainer), prefer the flag over the imperative API to keep the schedule declarative and reproducible.

Select the traceback or open the failed terminal, then run Denpex locally to see the initiating rank, collateral failures, exact fix, and verification command without uploading the log.

• An FSDP gradient reduce-scatter did not complete within the collective timeout. Like any collective timeout this names the ranks that were WAITING, not the one that failed to arrive, one rank left the collective early or died, and the rest report the timeout.
• The decisive evidence is the first log line that precedes "FSDP reduce_scatter backward hook timed out waiting for gradient sync" and differs from a healthy run.
• A nearby failure remains a competing hypothesis until a control separates configuration, capacity, transport, and hardware causes.