Docs/Error messages/An internal error occurred
Error messagesResources

An internal error occurred and the request could not be completed

BigQuery failed on its own side and says so with HTTP 500 and the reason internalError. Most of the time a retry with backoff works. If the same query fails the same way every time, the query is the trigger and the retry will not help.

An internal error occurred and the request could not be completed. This is usually caused by a transient issue. Retrying the job with back-off as described in the BigQuery SLA should solve the problem: https://cloud.google.com/bigquery/sla. If the error continues to occur please contact support at https://cloud.google.com/support. Error: 41416439

What BigQuery means

Something failed inside BigQuery, and it is not your SQL as far as BigQuery can tell. The API answers with HTTP 500 and the reason internalError. BigQuery’s documentation for that reason says to wait according to the back-off rules in the Service Level Agreement and try again, and to contact support if the error continues. It adds that you can reduce how often it happens by using reservations.

The message ends with a number. It is the only detail the error gives, so keep the whole message when you ask for help.

Most of the time the retry works. Sometimes it does not, because the query itself sets the error off. The example below fails the same way every time.

A query that always fails

SELECT 1 AS _FIELD_2, 2 AS _field_x

A dry run of this query returned the message at the top of the page, with the number 41416439. We repeated the dry run four times over about 50 seconds, waiting longer each time, and got the same message and the same number every time. The query has nothing wrong with its syntax or its tables.

Running the same query for real returned a normal error that names the problem:

Duplicate field '_field_2'

The reason for that one is invalidQuery, HTTP 400. SELECT STRUCT(1 AS _FIELD_2, 2) AS s behaves the same way: an internal error on the dry run, and Duplicate field 's._field_2' on a run. Both queries gave these results on 2026-11-04. BigQuery may have fixed this since, so read the example as one case of an internal error that a query causes.

The cause is a naming rule. BigQuery replaces a column name that starts with _field_ by _field_ and the column’s position. In the real run, _field_x in the second column became _field_2, which is the first column’s name _FIELD_2 when case is ignored. In SELECT 1 AS _field_x, 2 AS _FIELD_2 the first column was renamed to _field_1 and the query ran.

The fixes

  1. Retry with backoff. Wait, try again, and wait longer after each failure. This is what the message and the documentation ask for, and it is right for an error that comes and goes.
  2. If the same query fails every time, stop retrying. Take the query apart. Remove a column, a join or a function at a time until it runs, and the last thing you removed is the trigger. Then write that part another way.
  3. Contact support if it keeps happening. The message says so. Give them the whole message, the job ID and the query.

Retry only a query that is safe to run twice. A failed SELECT is. A script that failed part way has already done its earlier statements.

The API uses other reasons for the same kind of failure. These come from the BigQuery documentation:

ReasonHTTP codeMeaning
internalError500An internal error within BigQuery. Retry with backoff.
backendError500, 502, 503, 504The service is unavailable for a short time. Causes include a surge in demand, network problems and load shedding. Retry with truncated exponential backoff.
jobInternalError400The job was created, then failed with an internal error. Retry the job with a new job ID.
jobBackendError400The job was created, then failed with an internal error. Retry with a new job ID.

rateLimitExceeded is also temporary and is retried the same way. See Exceeded rate limits: too many table update operations.

Different errors that look like this one

An error that names a cause is not an internal error. Resources exceeded during query execution means the query was too heavy for its slot, and a retry will not change that. See Resources exceeded: allotted memory.

In Querylab.io

Querylab.io treats internalError, backendError, rateLimitExceeded and the three job... variants as temporary failures. When a run of a single SELECT fails with one of them, the status bar shows BigQuery’s message and offers Try again, which runs the query once more. The button does not appear for a script, a DML or DDL statement, or an error from a dry run. A dry-run error shows the message and has no one-click fix.