BigQuery insertId dedup is best-effort for 1 minute, not a guarantee

TL;DR

When building BigQuery streaming ingestion, generate stable insertIds for retries but never assume duplicates are gone: add a downstream dedup step (e.g. Omitting insertId entirely is the recommended way to get higher streaming ingest quotas; dedup strictly in your own pipeline instead.

Steps

1. When building BigQuery streaming ingestion, generate stable insertIds for retries but never assume duplicates are gone: add a downstream dedup step (e.g. ROW_NUMBER over the natural key in the query layer) for correctness. If throughput matters more than dedup, drop insertId for the higher quota. Agents writing streaming pipelines should document the dedup contract explicitly: best-effort at the wire, exact in the query.

When to use

You are seeing this: ROW_NUMBER over the natural key in the query layer) for correctness. Use this skill when you run into "BigQuery insertId dedup is best-effort for 1 minute, not a guarantee".

When not to use

If your error message or symptom does not match what is described above, this is probably not your fix. Search for your exact error text instead of forcing this one to fit.

Versions

No specific versions are mentioned in the source material, so treat the fix as generally applicable and check the examples against whatever you have installed.

Why this happens

The original report does not dig into a root cause. It documents the symptom and the fix that resolved it.

Source: Google Cloud docs (streaming data into BigQuery): supplying insertId gives best-effort deduplication for up to one minute only, and only maybe. If two rows share an insertId, it is nondeterministic which one survives, and Google can degrade dedup quality at any time to protect availability. It exists for retry scenarios, not as a guarantee. Omitting insertId entirely is the recommended way to get higher streaming ingest quotas; dedup strictly in your own pipeline instead.

Verified resolution: When building BigQuery streaming ingestion, generate stable insertIds for retries but never assume duplicates are gone: add a downstream dedup step (e.g. ROW_NUMBER over the natural key in the query layer) for correctness. If throughput matters more than dedup, drop insertId for the higher quota. Agents writing streaming pipelines should document the dedup contract explicitly: best-effort at the wire, exact in the query.

Source URL: https://docs.cloud.google.com/bigquery/docs/streaming-data-into-bigquery