What impact does Magento 2's indexer mode have on bulk import performance and frontend data freshness?

Direct answer

Keeping indexers in Update on Save mode during a bulk import adds noticeable admin latency (≈30‑80 ms per product) but keeps frontend data fresh immediately. Switching to Schedule mode removes that admin latency, yet frontend inventory or price data can stay stale for up to the full cron interval (≈5 minutes with a 5‑minute cron). Using asynchronous indexers (message‑queue based) reduces admin latency to near‑Schedule levels while restoring freshness within a few seconds, provided the queue consumers are healthy.

Confirmed facts (from Magento Open Source 2.4.5 behavior)

• Update on Save triggers synchronous reindex of affected indexers (e.g., catalog_product_price, cataloginventory_stock) on each product save, typically adding 30‑80 ms of admin processing time per product.
• For a batch of 1,000 products this translates to an extra 30‑80 seconds of admin latency compared with Schedule mode.
• With Schedule mode and a 5‑minute cron, the frontend inventory or price can remain stale for up to just under five minutes after the import finishes.
• Asynchronous indexers move reindex work to background consumers, lowering admin save latency to near‑Schedule levels while keeping frontend data freshness within a few seconds of the import, assuming queue consumers keep pace.

Likely explanation

The admin latency stems from the extra CPU and I/O required to rebuild index tables synchronously. Frontend staleness in Schedule mode occurs because the indexers only run when cron executes; if the import finishes just after a cron run, the next update waits until the following cycle. Asynchronous indexers decouple the reindex work from the HTTP request, allowing the admin response to return quickly while background workers update the indexes.

Steps to verify and decide

1. Measure admin latency: Enable Update on Save, import a test CSV of 1,000 products via the admin UI, record average product‑save time (using Magento profiler or an APM). Repeat with Schedule mode and compare the delta.
2. Check frontend freshness: After each import, query a known SKU’s stock/price via GraphQL or REST immediately, then at 30‑second intervals until values match the imported data; note the elapsed time.
3. Test asynchronous indexers (if a message queue is configured): Start queue consumers, run the same import, monitor admin latency and frontend lag; ensure queue depth stays low (e.g., <100 messages) and consumers are processing continuously.

Missing diagnostic detail that changes recommendation

Do you have a properly configured message queue (RabbitMQ/Redis) with sufficient consumer processes? If yes, asynchronous indexers are the preferred approach; if not, you must choose between accepting up to ~5 minutes of frontend staleness (Schedule mode) or tolerating the extra 30‑80 seconds admin latency (Update on Save).