Elasticsearch "search_phase_execution_exception" in CI

A search failed while executing across shards. The wrapper search_phase_execution_exception carries a more specific caused_by, most commonly that you sorted or aggregated on a text field with fielddata disabled, or sent a malformed query. Reading the nested reason is the whole fix.

What this error means

A search returns HTTP 400 with "search_phase_execution_exception" and a caused_by such as "Fielddata is disabled on [field] in [index]" or a query parsing error.

elasticsearch

{"error":{"type":"search_phase_execution_exception","reason":"all shards failed",
"caused_by":{"type":"illegal_argument_exception","reason":"Fielddata is disabled on
[name] in [products]. Text fields are not optimised for operations that require per-document
field data like aggregations and sorting ..."}},"status":400}

Common causes

Sorting or aggregating on a text field

Text fields have fielddata disabled by default; sorting or aggregating on one raises an error inside the search phase.

A malformed query body

A bad query clause or an unmapped field referenced in a sort or aggregation makes the shards fail during execution.

How to fix it

Aggregate on a keyword sub-field

Read the nested caused_by reason to see which field and operation failed.
Map a keyword sub-field (or use one) for sorting and aggregation.
Point the sort or aggregation at field.keyword instead of the analyzed text field.

Terminal

curl -X PUT "http://localhost:9200/products" -H 'Content-Type: application/json' -d '
{"mappings":{"properties":{"name":{"type":"text","fields":{"keyword":{"type":"keyword"}}}}}}'

Fix the query body

If the cause is a parse error, correct the query clause or reference a field that exists in the mapping.

How to prevent it

Always read the caused_by reason, not just the wrapper exception.
Use keyword fields for sorting and aggregations.
Validate queries and referenced fields against the mapping.