Skip to content

Common Mistakes & Anti-Patterns

When to Use

Use this when debugging search issues or reviewing a Search API implementation.

Decision

Issue Cause Solution
Search returns no results Index empty or not indexed drush sapi-i to index, drush sapi-s to check status
Results in wrong order Not sorted by relevance Add "Search: Relevance" sort descending in Views
Restricted content visible Content Access processor not enabled Enable Content Access processor on index
Facets show IDs not labels translate_entity not enabled on facet Enable processor on the facet
Taxonomy terms wrong after edit Reference tracking disabled or not reindexed Enable track_changes_in_references, reindex
Slow search on DB backend Too many items for database Switch to Solr
Slow COUNT queries on MySQL InnoDB COUNT limitation Use Mini pager or skip result count
Memory exhaustion during indexing Batch size too large Reduce cron_limit to 10-25
Processors not taking effect Old data in index drush sapi-r && drush sapi-i
Stemmer conflicts on Solr Duplicate processing Disable Tokenizer, Stemmer, Stopwords, Ignore case for Solr
Core Search module conflicts Both modules running drush pm:uninstall search
Highlight adds latency Processor too expensive Disable or use search_api_skip_processor_highlight tag
Search returns all content unfiltered Missing fulltext search exposed filter Add "Search: Fulltext search" filter to View

Pattern

Debugging checklist: 1. Check index status: drush sapi-s — is the index fully indexed? 2. Check server availability: Admin → Server page → "Server connection" status 3. Check processors: Are the right processors enabled for your backend? 4. Check Views sort: Is "Search: Relevance" added as sort criteria? 5. Check fulltext filter: Is the fulltext search exposed filter present? 6. Check field types: Are facet fields String (not Fulltext)? 7. Test reindex: drush sapi-r && drush sapi-i 8. Check permissions: Test as anonymous user 9. Check logs: drush ws --type=search_api

Common Mistakes

  • Wrong: Not creating key pages as nodes → Right: Custom routes/blocks are not indexed. Pages that matter for search should be nodes.
  • Wrong: Expecting immediate results after content edit when index_directly: FALSE → Right: Changes appear after next cron run.
  • Wrong: Not reindexing after processor changes → Right: Always reindex after changing processors.

See Also