Elasticsearch "version_conflict_engine_exception" in CI

Elasticsearch rejected a write because the document version it was given did not match the current version. This optimistic-concurrency guard fires when two writes race, when a create request targets an id that already exists, or when a test reuses a stale sequence number.

What this error means

An index, update, or delete returns HTTP 409 with "version_conflict_engine_exception", mentioning the current and required primary term or sequence number.

elasticsearch

{"error":{"root_cause":[{"type":"version_conflict_engine_exception","reason":
"[1]: version conflict, required seqNo [5], primary term [1]. current document has
seqNo [7] and primary term [1]"}],"type":"version_conflict_engine_exception"},"status":409}

Common causes

Concurrent writes to the same document

Two updates race on the same id; the second carries a sequence number that is already stale, so it is rejected.

A create request hit an existing id

An op_type=create (or _create) request against an id left over from a previous run conflicts with the existing document.

How to fix it

Reset state and avoid id reuse between runs

Delete or recreate the index so leftover ids do not collide.
Use unique ids per test, or let Elasticsearch auto-generate them.
Refresh between dependent writes so reads see the latest version.

Terminal

curl -X DELETE "http://localhost:9200/orders"
curl -X PUT "http://localhost:9200/orders"

Retry on conflict for concurrent updates

For update-by-query or update requests, set retry_on_conflict so a losing writer retries against the current version instead of failing.

Elasticsearch API

POST /orders/_update/1?retry_on_conflict=3
{"doc":{"status":"paid"}}

How to prevent it

Start each run from a clean index so document ids do not carry over.
Use auto-generated ids or per-test unique ids.
Set retry_on_conflict on updates that can race.