dbt cloud job failing with out of memory error

TL;DR

The job process ran out of memory, usually because of too many threads on a small runner or one giant model. Find the failing step in the run logs, lower the thread count in the job settings, split models across jobs, and check for runaway queries (like accidental cartesian joins) inflating row counts.

Error

dbt cloud job failing with out of memory error

Steps

1. Open the failed run's logs in dbt Cloud and find the step where memory ran out. Expected: you know which model or phase failed.
2. Check the model's row counts and its SQL for fanout (joins multiplying rows). Expected: you rule out a runaway query first, since no setting fixes a broken join.
3. Lower the thread count in the job's settings. Expected: fewer concurrent models means lower peak memory.
4. Split the job: move heavy models to a separate job or schedule. Expected: each job's peak memory drops.
5. Rerun the job and watch memory. Expected: it completes without the OOM error.

When to use

• A dbt Cloud job fails with out of memory.
• The failure started after adding models or raising threads.

When not to use

• The warehouse reports the out-of-memory (scale the warehouse, not the dbt job).
• The job fails fast with a config error (not a resource problem).

Tool compatibility

• dbt Cloud (all tiers). Thread settings live in the job definition.

Variant phrasings

Job killed, exit code 137

The container OOM-killer fired; same fix: less concurrency, smaller scope.

OOM only on the full-refresh job

Full refreshes materialize whole tables at once; they need more headroom than incremental runs.

Why it happens

dbt holds compiled artifacts, results, and adapter state in memory, multiplied by thread count. More threads and bigger models raise the peak until the container limit kills the process.

Edge cases

• --store-failures on huge test failures can spike memory; scope the failing tests first.
• Docs generation on giant projects is also memory-hungry; run it as its own job.
• If OOM persists at minimal threads, the runner size itself is the constraint; contact support about larger runners.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstvCgSce3-hfQU-1PR3ySNQ

Resolved from the public thread: https://vectle.com/posts/pstvCgSce3-hfQU-1PR3ySNQ

Thread query: dbt cloud job failing with out of memory error

TL;DR: The job process ran out of memory, usually because of too many threads on a small runner or one giant model. Find the failing step in the run logs, lower the thread count in the job settings, split models across jobs, and check for runaway queries (like accidental cartesian joins) inflating row counts.