Skip to content

Indexing Performance

When to Use

When optimizing how fast content gets indexed, especially for large sites or initial indexing.

Decision: Batch Size Tuning

Scenario Recommended Batch Size Why
Simple nodes (title + body) 100-200 Low memory per item
Complex nodes (paragraphs, many references) 10-25 High memory per item (200MB+ per batch)
Initial full index Use drush, not cron More control over duration
Ongoing incremental 50-100 via cron Balance freshness vs load

Set in Index → Edit → "Cron batch size" or via drush --batch-size.

Decision: Cron vs Drush

Method Use Case Command
Cron Incremental indexing of content changes Automatic (set cron_limit on index)
Drush Initial indexing, reindexing, bulk operations drush sapi-i --batch-size=50 --time-limit=300
Solr parallel indexing Large sites (10K+ items) on Solr drush search-api-solr:index-parallel --threads=8 --batch-size=100
search_api_fast Large sites (10K+ items) on a non-Solr backend Spawns parallel workers across CPU cores

Pattern: Drush Indexing Commands

# Index all pending items
drush sapi-i

# Index specific index only
drush sapi-i my_index

# With batch size and time limit
drush sapi-i --batch-size=25 --time-limit=600

# Mark everything for reindex (doesn't delete from backend)
drush sapi-r my_index

# Clear index completely (deletes from backend)
drush sapi-c my_index

# Rebuild tracking info
drush sapi-rt my_index

Pattern: Parallel Indexing

On Solr, use the parallel indexing command that ships with search_api_solr — no extra module: 

drush search-api-solr:index-parallel my_index --threads=8 --batch-size=100

On a non-Solr backend, search_api_fast spawns parallel Drush workers: 

composer require drupal/search_api_fast
drush search-api-fast:index my_index
# or with the alias:
drush sapi-fast my_index

Worker count is configuration, not a command option — there is no --workers flag. Set index_workers (default 8) in search_api_fast.performance: 

drush config:set search_api_fast.performance index_workers 4

Pattern: Solr Index-Only Mode

For maximum query performance on Solr, skip entity loads entirely:

  1. Server config → Enable "Retrieve result data from Solr"
  2. Views query settings → Enable "Skip item access checks" (only for fully public content)
  3. Ensure all displayed fields are in the Search API index

Result: Solr returns field data directly — no database queries for entity loading.

Pattern: Cron Frequency

Site Size Cron Interval Batch Size
Small (<10K items) 15 minutes 100
Medium (10K-100K) 5 minutes 50
Large (100K+) 1-2 minutes 25

Common Mistakes

  • cron_limit = -1 on large indexes — Indexes all items at once. Causes memory exhaustion on complex entities.
  • cron_limit = 0 — Disables cron indexing entirely. Items never get indexed via cron.
  • Not using drush for initial indexing — Cron indexing large datasets takes days. Use drush with --time-limit for controlled batch indexing.

See Also