Skip to content

Host RAM OOM-Kill of DataLoader Workers Misread as GPU OOM

The kernel oom-killer (or cgroup limit) kills python workers whose RSS grew unbounded. Shard caches, tokenized-in-RAM datasets, prefetch queues. Surface strings ("Out of memory", "Killed") route engineers to GPU OOM playbooks that cannot help.

Quick answer

The kernel oom-killer (or cgroup limit) kills python workers whose RSS grew unbounded. Shard caches, tokenized-in-RAM datasets, prefetch queues.

Memory#host-oom#oom-killer#dataloader#cgroup#rss#not-gpu-oom

What this failure is

Host RAM OOM-Kill of DataLoader Workers Misread as GPU OOM is a Memory failure seen during ML training runs. The kernel oom-killer (or cgroup limit) kills python workers whose RSS grew unbounded. Shard caches, tokenized-in-RAM datasets, prefetch queues. Surface strings ("Out of memory", "Killed") route engineers to GPU OOM playbooks that cannot help. Common tags: Host Oom, Oom Killer, Dataloader, Cgroup.

Live diagnosis, no signup

Is this what broke your run? Paste your log.

You're reading about Host RAM OOM-Kill of DataLoader Workers Misread as GPU OOM. 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)

Unbounded in-worker cache (dict/LRU without cap). Copy-on-write defeated by python refcounting on forked datasets. Rank0 auxiliary services inflating one pod. 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

  • DataLoader worker (pid N) is killed by signal: Killed
  • kernel: Out of memory: Killed process (python3) with huge anon-rss
  • GPU memory never near capacity at kill time

Common symptoms and what they mean

SymptomWhy it happens
exitcode -9 on a worker; torchrun ChildFailedError wrapperunbounded in-worker cache (dict/LRU without cap)
RSS staircase growth per epoch in node metricscopy-on-write defeated by python refcounting on forked datasets
no torch.cuda.OutOfMemoryError anywhererank0 auxiliary services inflating one pod

Which systems are affected

  • num_workers>0 pipelines with per-worker caches
  • pin_memory with large prefetch_factor
  • k8s pods with tight memory limits

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: exitcode -9 on a worker; torchrun ChildFailedError wrapper
  • ✓Verified signal present: RSS staircase growth per epoch in node metrics
  • ✓Verified signal present: no torch.cuda.OutOfMemoryError anywhere
  • ✓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 analysis

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

Related failures to investigate next

Cpu Offloading Overhead often co-occurs with this failure because both share underlying assumptions about memory; if you fix this entry but still see drift, validate the related primary cause next. cpu offloading overhead →
Gradient Checkpointing Tradeoff often co-occurs with this failure because both share underlying assumptions about memory; if you fix this entry but still see drift, validate the related neighboring cause next. gradient checkpointing tradeoff →
Cuda Oom often co-occurs with this failure because both share underlying assumptions about memory; if you fix this entry but still see drift, validate the related neighboring cause next. cuda oom →

Root cause

  • unbounded in-worker cache (dict/LRU without cap)
  • copy-on-write defeated by python refcounting on forked datasets
  • rank0 auxiliary services inflating one pod

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.

We send a single-use code tied to that address. Static provider and TLD rules do not reject valid addresses. Account trust determines the benefit after signup.

References

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 freeBrowse all failures

Related Memory errors

Activation Memory Spike

Memory · high

Memory Summary Tool Usage

Memory · low

PyTorch Caching Allocator Debug

Memory · low

CPU RAM Exhaustion During Training

Memory · high