Is this a hardware or a software issue?

Hardware/Network failures can look like either layer depending on the symptom. The discriminator is the resolution of the symptom to the recommended fix: if applying the steps under "How to fix it" eliminates the symptom, this is a software/configuration class issue regardless of the hardware it surfaced on. Repeat the fix on a fresh node to confirm.

How long does the fix typically take?

Most resolutions from this entry take 10 to 30 minutes of engineer time once the diagnosis is confirmed. Long fixes come from skipping the diagnosis step and chasing apparent symptoms (validation drift, memory growth, slow throughput) instead of the actual root cause listed above. Apply the fix in a small experiment first, then roll it forward.

Does this affect inference or only training?

Primarily this is a hardware/network failure observed during training, but the same misconfiguration can leak into downstream evaluation and serving. Treat the prevention checklist as a guardrail: any pipeline that touches a trained checkpoint should re-validate the failure mode before promoting the model.

What's the cheapest way to confirm this is the problem?

Cheapest signal: reproduce the symptom on a 1-GPU toy run with the same configuration. If the toy run exhibits the same behavior, the cause is the configuration rather than the cluster. If only the multi-GPU run is affected, the failure is a distributed interaction, so Megatron-LM is the most likely layer to inspect.

Can the Denpex platform detect this automatically?

Yes. Denpex correlates the symptoms in this entry against your training logs in real time and surfaces the failure taxonomy, root cause, and exact remediation command on the console. Three full diagnoses are included free during signup, and the encyclopedia entry you are reading is generated from the same signal library that powers the platform.

What related failures should I check next?

Closely related failures cluster around #NCCL Deadlock. After applying this fix, validate against the related entries below to confirm no overlapping cause is masking as this one. Engineers routinely discover that the dominant failure mode is actually a second, sibling failure compounded on top.

Will this recur after the fix?

Applied once with the prevention checklist, this failure does not recur under normal operation. The recurring variant most engineers encounter is the re-emergence of the same symptom on a different model or stack after NCCL Deadlock is upgraded. Pin compatible versions and rerun the prevention smoke test on every dependency change.

Is there a fingerprint I can monitor?

Yes. Training with Mixture of Experts (MoE) hangs sporadically when scaling to multiple nodes. is the highest-signal fingerprint. Wire a metric or alert on that signal (or its absence after the fix) so you can detect the recurrence automatically. The Denpex platform encodes this exact fingerprint and surfaces it directly in the dashboard.

What is the official documentation that covers this?

Reference documentation is listed in the References section at the bottom of this entry. The most authoritative URLs are the upstream framework docs and the issue trackers cited; the Denpex entry synthesizes both into the executive, engineer, and on-call perspectives so each audience can act without reading every linked source.

Does this entry include working code?

Yes. The "Code Examples" section contains copy-paste-ready snippets that target the most common stacks (PyTorch, Hugging Face Trainer, DeepSpeed, Megatron-LM, vLLM/Transformers). Each snippet is annotated with why it works so you can adapt the pattern to your own framework without losing fidelity.

MoE Expert Parallelism Comm Overlap Deadlock

Q: What does this failure look like in production?

In production this failure typically appears during the first tens-to-hundreds of training steps: loss curves become non-monotonic, validation metrics plateau above the baseline, and downstream checkpoints either diverge or fail to improve. Engineers frequently misread it as data drift when the underlying mechanism is the moe overlap scheduler uses a shared cuda event as a 'baton' to sequence communications across nodes. due to timing variations, ranks can launch overlapping nccl collectives in different orders (e.g., rank 0 launches all2all_a then all2all_b, while rank 1 launches all2all_b then all2all_a). nccl mandates that all ranks participating in a collective must submit the collective calls in the exact same sequence. the mismatch causes a cross-rank nccl deadlock.. Validating against the symptom list in this entry is the fastest way to confirm the diagnosis.

Q: How does this scale from 1 GPU to 1000 GPUs?

At single-GPU scale the symptom usually looks like slow convergence or a single anomalous loss value. At 64+ GPU scale, distributed interactions amplify the moe overlap scheduler uses a shared cuda event as a 'baton' to sequence communications across nodes. due to timing variations, ranks can launch overlapping nccl collectives in different orders (e.g., rank 0 launches all2all_a then all2all_b, while rank 1 launches all2all_b then all2all_a). nccl mandates that all ranks participating in a collective must submit the collective calls in the exact same sequence. the mismatch causes a cross-rank nccl deadlock. through collectives, and the failure can manifest as a full job abort. Always reproduce the fix at the smallest cluster configuration that still exhibits the symptom before promoting to production scale.

The MoE overlap scheduler uses a shared CUDA event as a 'baton' to sequence communications across nodes. Due to timing variations, ranks can launch overlapping NCCL collectives in different orders (e.g., Rank 0 launches All2All_A then All2All_B, while Rank 1 launches All2All_B then All2All_A). NCCL mandates that all ranks participating in a collective must submit the collective calls in the exact same sequence. The mismatch causes a cross-rank NCCL deadlock.

Quick answer

The MoE overlap scheduler uses a shared CUDA event as a 'baton' to sequence communications across nodes.

Symptom: NCCL WARN [send] via NET/IB/0/GDRDMA failed
Root cause: The MoE overlap scheduler uses a shared CUDA event as a 'baton' to sequence communications across nodes. Due to timing variations, ranks can launch overlapping NCCL collectives in different orders (e.g.
Recommended fix: --overlap-moe-expert-parallel-comm
How Denpex helps: Denpex matches MoE Expert Parallelism Comm Overlap Deadlock across every rank in a distributed run and reports which rank failed first, so you act on the initiating node instead of the loudest one.

Hardware/Network#NCCL Deadlock

What this failure is

MoE Expert Parallelism Comm Overlap Deadlock is a Hardware/Network failure seen during ML training runs. The MoE overlap scheduler uses a shared CUDA event as a 'baton' to sequence communications across nodes. Due to timing variations, ranks can launch overlapping NCCL collectives in different orders (e.g., Rank 0 launches All2All_A then All2All_B, while Rank 1 launches All2All_B then All2All_A). NCCL mandates that all ranks participating in a collective must submit the collective calls in the exact same sequence. The mismatch causes a cross-rank NCCL deadlock. Common tags: NCCL Deadlock.

Live diagnosis, no signup

Is this what broke your run? Paste your log.

You're reading about MoE Expert Parallelism Comm Overlap Deadlock. Paste your own crash log or traceback below and get the real root cause for YOUR run, not this generic entry. No account, no card. Logs are masked at ingress and never saved to account history.

training_logs.txt

No log to hand? Try one:

3 free diagnoses/day

Want 14 days on the Scale plan?

Request an evaluation code. A verified workplace organization activates up to 50 diagnoses a day, alerts, history, and follow-up questions. No credit card or automatic subscription.

Evaluate one incident

Why it happens (the mechanism)

Because it happens asynchronously on separate CUDA streams, the stack trace usually points to a completely unrelated PyTorch operation (like a linear layer forward pass) that is waiting for the default stream to sync.

What you'll observe

NCCL WARN [send] via NET/IB/0/GDRDMA failed
NCCL INFO Call to ncclAllGather failed
Hangs during MoE routing or expert collective communication

Common symptoms and what they mean

Symptom	Why it happens
Training with Mixture of Experts (MoE) hangs sporadically when scaling to multiple nodes.	The MoE overlap scheduler uses a shared CUDA event as a 'baton' to sequence communications across nodes. Due to timing variations, ranks can launch overlapping NCCL collectives in different orders (e.g., Rank 0 launches All2All_A then All2All_B, while Rank 1 launches All2All_B then All2All_A). NCCL mandates that all ranks participating in a collective must submit the collective calls in the exact same sequence. The mismatch causes a cross-rank NCCL deadlock.
Only occurs when `--overlap-moe-expert-parallel-comm` is enabled.	The MoE overlap scheduler uses a shared CUDA event as a 'baton' to sequence communications across nodes. Due to timing variations, ranks can launch overlapping NCCL collectives in different orders (e.g., Rank 0 launches All2All_A then All2All_B, while Rank 1 launches All2All_B then All2All_A). NCCL mandates that all ranks participating in a collective must submit the collective calls in the exact same sequence. The mismatch causes a cross-rank NCCL deadlock.

Symptom

Why it happens

Training with Mixture of Experts (MoE) hangs sporadically when scaling to multiple nodes.

Only occurs when `--overlap-moe-expert-parallel-comm` is enabled.

Which systems are affected

Megatron-LM
NCCL
CUDA Streams

How to confirm this is the problem

Use this checklist to test the hypothesis against a small reproduction. No single line proves the root cause, so preserve the preceding events and compare one variable at a time.

✓Run with `NCCL_DEBUG=INFO` and `NCCL_DEBUG_SUBSYS=COLL` to trace the sequence of collectives.
✓Look for mismatching collective sequence numbers across different ranks right before the hang.
✓Temporarily disable `--overlap-moe-expert-parallel-comm` to confirm it is the root cause.

Searchable error signature

search key

NCCL WARN [send] via NET/IB/0/GDRDMA failed
NCCL INFO Call to ncclAllGather failed

Use this text as a lookup key in logs and upstream issue trackers. It is not presented as a captured customer log. Confirm the cause from your own preceding events, versions, configuration and the cited references.

The fix and the prevention pattern

The root cause is on this page and stays free. A free account adds the exact remediation steps, saved history, and the fix on every entry in the encyclopedia.

No credit card. Daily allowance follows verified trust tier. Instant access.

Diagnose this failure in VS Code

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.

Install the free VS Code extension

NCCL errors in context

NCCL is where a distributed job reports failure, which is not the same as where it failed. The hub lists every common NCCL error next to what it actually indicates, and the environment variables that tell them apart.

Compare every nccl error side by side

Root cause

The MoE overlap scheduler uses a shared CUDA event as a 'baton' to sequence communications across nodes. Due to timing variations, ranks can launch overlapping NCCL collectives in different orders (e.g., Rank 0 launches All2All_A then All2All_B, while Rank 1 launches All2All_B then All2All_A). NCCL mandates that all ranks participating in a collective must submit the collective calls in the exact same sequence. The mismatch causes a cross-rank NCCL deadlock.

The fix and how to prevent it

Evaluate Denpex on your own logs

Request a Scale evaluation code. A verified workplace organization activates 14 days with up to 50 diagnoses a day. Every account keeps its current diagnosis allowance and gets a verification path. No card or automatic subscription.

References

↗GitHub Issue #1810: Overlap MOE Expert Parallel Comm Deadlock

Don't just read the fix, diagnose your run

The encyclopedia tells you what went wrong. Denpex tells you what went wrong in YOUR training run. With your logs, your config, and your stack.

Diagnose your logs free Browse all failures

Related Hardware/Network errors

NVLink Fatal Error Xid 74 with Fabric Manager Crash

Hardware/Network · critical