RoCE QP Timeout During Distributed Checkpoint Save
NCCL queue-pair creation times out specifically during distributed checkpoint save at large scale on RoCE, while normal training collectives succeed.
NCCL queue-pair creation times out specifically during distributed checkpoint save at large scale on RoCE, while normal training collectives succeed.
What this failure is
RoCE QP Timeout During Distributed Checkpoint Save is a Communication failure seen during ML training runs. NCCL queue-pair creation times out specifically during distributed checkpoint save at large scale on RoCE, while normal training collectives succeed. Common tags: Nccl, Roce, Infiniband, Checkpoint.
Is this what broke your run? Paste your log.
You're reading about RoCE QP Timeout During Distributed Checkpoint Save. 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.
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.
Why it happens (the mechanism)
The gather/star communication pattern of checkpoint save stresses the fabric differently than ring all-reduce. RoCE v2 congestion or PFC/ECN misconfiguration at scale. Per-rank metadata collection overloads rank 0 / saturates QPs. Taken together, these mechanisms explain why the failure is reproducible, why it tends to surface on specific workloads or scales, and why generic mitigation attempts often fall short without addressing the underlying cause.
What you'll observe
- Training runs fine but dcp.save() fails at scale (e.g. 8 nodes / 64 GPUs)
- Smaller node counts (2/4/6) work
- Failure is in the checkpoint gather, not the training step
Common symptoms and what they mean
| Symptom | Why it happens |
|---|---|
| ibv_modify_qp failed with error Connection timed out errno 110 | The gather/star communication pattern of checkpoint save stresses the fabric differently than ring all-reduce |
| RuntimeError: NCCL Error 2: unhandled system error | RoCE v2 congestion or PFC/ECN misconfiguration at scale |
| Hang then timeout during torch.distributed.checkpoint save | Per-rank metadata collection overloads rank 0 / saturates QPs |
Which systems are affected
- RoCE v2 clusters at 32+ GPU scale
- Distributed checkpoint (dcp.save) with gather to rank 0
- H800/H100 nodes on bonded RoCE NICs
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.
- ✓Reproduce the failure from a clean checkpoint/seed: the symptom must appear without warm-up state from a previous run.
- ✓Verified signal present: ibv_modify_qp failed with error Connection timed out errno 110
- ✓Verified signal present: RuntimeError: NCCL Error 2: unhandled system error
- ✓Verified signal present: Hang then timeout during torch.distributed.checkpoint save
- ✓A targeted fix from the "How to fix it" section eliminates or substantially reduces the symptom within one validation pass.
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.
Sign up free. Unlock the full analysisNo 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 extensionNCCL 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 sideRelated failures to investigate next
Root cause
- The gather/star communication pattern of checkpoint save stresses the fabric differently than ring all-reduce
- RoCE v2 congestion or PFC/ECN misconfiguration at scale
- Per-rank metadata collection overloads rank 0 / saturates QPs
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.
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.