The speed of the homepage and products is not inconsistent with Checkout. Public pages often respond from the page cache or CDN, but Checkout must process the session, box, address, submission, tax, discount and payment for the same user. This dynamic path is directly dependent on PHP, database and external services, and does not measure the good score of the homepage.
Quick answer:In DevTools, specify the delay when opening a page, whether it's a Checkout update request, order registration, or port transfer. Match the delay request with the same PHP/WooCommerce log, Query, and external call timestamp. Do not cache the Checkout page and do not allow the increase in timeout to serve as a service finding.
Why does Page Cache hide the difference?
A cached output page may be delivered without full execution of WordPress, while Checkout will consume for each actual request worker. An anonymous test of the host page does not prove the dynamic capacity of the server. Compare the cache header, cookie, and origin time in both directions.
What time is slowdown coming?
| Hold on a second. | The main candidates | Witness. |
|---|---|---|
| Early opening. | PHP, templates, sessions and queries | TTFB and profiler |
| Change of province/address | Ajax, shipping and tax. | Network waterfall |
| Select the method of transmission. | Recomputing and API provider | Request count and external log |
| Order recording | Validation, write, entity and hook | Ajax response and PHP log |
| Transfer to the payment gateway. | API, DNS, TLS and timeout | Gateway log and request ID |
Use DevTools on a controlled purchase.
Activate the Preserve log and separate document, XHR/Fetch and redirects. Enter status, duration, initiator and insensitive response. If the request is delayed, the backend or network is created; if the response is fast but the spinner remains, check the JavaScript and interface handler. Do not click on the payment multiple times.
Separate browser time and server time
In Network, queueing, DNS, connect, TLS, waiting, and download are different components. The above TTFB can be proxy/PHP or network waiting; server timing and upstream logging help to separate. If the server responds in 200 milliseconds but the browser is waiting for a few seconds for the handler, increasing the PHP worker is the wrong path.
Conversely, if JavaScript sends a quick request and the response is delayed, the browser flame chart does not show the cause of the backend. Create a shared request ID or timestamp to connect the browser trace, Nginx, PHP, and gateway.
Repeated requests for Update Checkout
Changing the address and method of sending can start the total calculation. An inappropriate script or hook may create a loop and send multiple requests behind the scenes. Compare the number, initiator, and unsensitive payload.
Transport and taxes
The sending provider may call an external API for any change in the field. Record connect/read time, timeout, call number, and fallback. The sending rate cache must be valid based on origin, destination, weight, and conditions; incorrect reuse can indicate the wrong price. Do not disable TLS validation for less than a few hundred milliseconds.
Gate and sidings.
When you place an order, the payment gatewayway, anti-heart, text, CRM or ERP may run synchronous. A slow provider keeps the entire response. Check the request ID, DNS, TLS, and status. Timeout is not necessarily a complete failure of the transaction; inquire about the previous status before retrying to avoid making a repeat payment.
Session and storage basket
Checkout requires a steady session. A cookie with an incorrect domain/HTTPS, storage, eviction or multiple incompatible nodes can cause time and error. Test with a guest and member, a new browser, and the final hostname. Do not register or publish the amount of cookie. Clearing all sessions will disrupt active user purchases.
Price personalization and trade rules
User role prices, major discounts, purchase limits, loyalty scores, and multi-store availability are usually re-evaluated in Checkout. If each rule runs separately for each line item Query or API, the product page caches up quickly but enlarges the cart.
For speed, do not delete price or item validation. The resulting cache must cover the entire effective user, product, number and time of validation and will be invalidated after the change.
Query and Lock when you place an order.
Calculating price, coupon, inventory and order creation has multiple read-write options. Heavy reporting, inventory importing or long transaction can create lock and I/O. See slow log and controlled Query Monitor in the same space. Test index and delete metadata with only backup, plan and clone.
PHP-FPM and the Worker line
When public pages are cached, the worker shortage is not visible until Checkout. Monitor active/idle worker, queue, ceiling access, and memory of each process. Increase worker without sufficient RAM creates OOM or swap. Assign the request time to component, and then adjust capacity.
Action Scheduler and Order Hooks
Some operations need to be in the background, but the plugin may do heavy work or consume a row of resources before responding. Check hook, time, and exception. Transferring financial operations to async without design is dangerous for idempotency and mid-level; the boundaries of transaction and customer experience should be clear.
Compare the differences between payment and delivery methods
Compare methods in a valid test environment to a fixed address and a box. If only one gateway is slow, a public query is less likely. If all methods are slow before the payment gatewayway is selected, calculating total, session, or PHP is more important.
Do not turn off the active production plug-in for customer testing. Have a staging or maintenance plan and make sure it does not create actual testing, email, inventory or fulfillment.
Do not cache the checkout.
HTML Checkout contains user-dependent data, nonce, total, and methods. Public cache may appear to be fast but spoils the cart, price, or security. Bypass should cover necessary paths and cookies. Object cache is different from page cache and can only be used with proper invalidation and separation.
A comparison that reveals the cause
- Record the hot and cold cache product page.
- Open the checkout with a simple product and a fixed address.
- Take each Ajax from the document separately.
- Compare the transmission and port in the controlled environment.
- Put PHP, Query and external call in the same timeline.
- Just change one component and repeat the scenario.
Order of correction
- Fix PHP/JavaScript errors and repeated requests.
- Slow down the API and adjust the timeout/retry.
- Change the query and hook to the original owner.
- Set the PHP array and database after measurement.
- Keep the public pages cache and verify checkout bypass.
- Test payments, callbacks, email and inventory end-to-end.
The measure of success.
- P50/p95 opening and checkout update
- POST time of order registration and port transfer
- 4xx/5xx rate and timeout
- PHP queue and query time
- Total correctness, shipping, coupon and tax
- One order and one transaction for every legitimate effort.
Common Mistakes
- Checkout comparison to the homepage is cached.
- Page caching Checkout
- Increased timeout and worker unrecognized
- Multiple push of the button.
- Turn off TLS or WAF.
- The door is off on the production in the middle of the purchase.
- Speed test without order accuracy control.
Related guide
If the request finally responds, but it's late, the guideslowdown Checkout and the Commerce.If the spinner stays or the order is made without redirect,- I'm getting the order.Follow him.
When do you need special assistance?
If a delay occurs only with actual payment, concurrent traffic, or a delivery method, the production test can create a faulty order.Support for the WooCommerce storeIt can check the network, PHP, Query and external API with an order ID in a single trace.
Common Questions
Is a CDN useful for checkout?
For static assets yes, but dynamic processing and APIs remain in origin and personal HTML should not be publicly cached.
Why is it slowed down by just choosing one province?
Check the same destination's sending/taxing methods and API in terms of number and latency.
Redis solves the problem?
Only if repeat database reading plays a significant role, latency does not solve the Ajax gate or loop.
Why can't the site manager see slowdown?
The address, method of sending, cache, session or user role is different.