thanks for your excellent work and this fantastic plugin.

There is a problem if using W3 with HTTP/3: there is no HTTP_HOST variable for nginx or in other words, there won’t be a HOST header in the request.

With this, your plugin will generate a page_enhanced cache file for H2, and another one for H3… difference is just the non-existence of the “HOST” domain…

• page_enhanced/URL -> for HTTP3
• page_enhanced/HTTP_HOST/URL -> for HTTP2 and others

It is a problem with W3 cache using HTTP/3, since with a HTTP3 configuration, there will be always browsers and bots asking for HTTP/2 also, of course, and there is going to be two different cache paths, one for each protocol.

It is possible to pass HOST variable to fastcgi (php) with HTTP3 using $host (nginx internal) instead of $http_host (header) and everything will be ok….

But if I do this way, your nginx.conf rules won’t run since they are checking files and rewriting by using $http_host variable, which is going to be empty, so will look for “page_enhanced/URL” and it won’t exist.

A suggestion could be for you to use $host variable in the nginx.conf rules for everything, and let user pass $host to php through fastcgi params… but we may consider $host variable won’t include TCP port….

With HTTP/3, nginx admins are almost forced to use $host instead of $http_host everywhere, since there won’t be a Host header, but a authority header….

Regards
Gino

Today, a Woocommerce site was misbehaving – 503 responses on WP login and many pages. On checking I found that the site was using 220,000 files on an allcoation of 200,000. It’s a Woocommerce site with around 4,000 products and it’s run well for years.

As a temporary measure, I deleted the cache file on CPanel – and the file count became 128,000. Soon though it was back to 220,000 so I had to disable the plugin and look for another.

Never seent his before.

Hi, I am having a caching issues that I believe are related to the Redis object cache. It shows up consistently after awhile on normal site pages but more drastically on woo commerce related pages. I have a website with around 2000 products and 80,000ish variations. So my sitemap is quiet big (which I supplied to the plugin also) I’ve noticed having this many variations while using woo commerce results in a large database ( my post meta table is over a million rows that I cant really clean because the bulk of it is from Woo commerce when it created my product variations. This large db while ugly does not seem to be effecting load like crazy according to Query Monitor Plugin. )

I believe my issue could be a mix of Redis and PHP settings on the server not allowing enough resources to be allocated and my configuration of the plugins settings (Browser/Page/Object tabs).

I have attempted some troubleshooting and tweaks and I believe that I have the object cache running because

1. if I’m logged out and I visit a page twice (not just refresh but leave and come back) it’ll load instantly the second time around.
2. It also works If I access a product page on PC1 and then access it again on a different PC right away, it will load fine most of the time. (sometimes PC2 will still experience long load time. usually always the document and waiting for server response time.)

This ‘test’ I do also only works for a few minutes because after ~5 minutes, maybe less I will experience slow load times on all of my pages and the network inspect during this load shows the document file primarily taking up the bulk of that time Waiting for server response.

It feels like my objects are being uncached even though the message by W3 at the bottom of the page shows around the same range of objects being cached on this page during slow and fast times. (this is typically on the Woocommerce pages but also home)

The Home page can also experience the long load times on the first visit if it hasn’t been visited recently. (but longer than the shop or product page) I’ve noticied that The W3 banner at the bottom will show a wide range of numbers here sometimes showing 48 / 792 object caching or 582 / 1493 but load some what quickly both times, does this number not really matter or do I not understand it?

It might be my lack of cache knowledge but is this supposed to ever fill up or stay at a consistent number? seems to move around. I know this space will give a warning too for blocked like when you’re logged in but I see no rejected when logged out so its not that at least.

Brief Settings (maybe there’s a better way to export these):

• Page Cache – Enabled – Redis
• Minify – Disabled
• Opcode Cache – Enabled – Opcode: Zend Opcache
• Database Cache Disabled
• Object Cache – Enabled – Redis
• Browser Cache – Enabled

Page Cache Settings

• General block is almost all enabled, “Cache URI’s with query string variables” isn’t enabled. also don’t cache pages for logged in users is enabled as well.
• Cache Preload = Automatically Prime the page cache,
• Updated interval = 900 seconds & Pages Per = 20 ( I do not understand what this is doing exactly and what combo of numbers helps here should I lower the interval and increase the pages per? )
• sitemap url is given and preload upon publish is checked.
• Late Initialization is unchecked ( would enabling this help )
• Persistent Connection is checked also.
• Connection Timeout + retry interval + timeout is all 0.
• Max Lifetime of cache objects = 3600 seconds (is this the expiry that I experience? does this mean it cleans everything and no pages load fast without a prior visit again? or does the Updated interval cycle through my pages and do this? but it wouldn’t complete in time if the expiry is that low. correct?)
• Never cache the following pages: wp-.*\.php & index\.php
• Rejected cookies ( woocommerce_items_in_cart, woocommerce_cart_hash, wp_woocommerce_session_, wordpress_logged_in )

Object Cache Settings

• Enabled Persistent connection
• Connection timeout, interval, read timeout are all 0
• Default lifetime of cache objects 180 seconds (again same question as above, and by larger the cache they mean more ram usage?)
• Garbage collection interval = 3600 seconds (This was greyed out in the page cache but not object cache, is this setting important?)
• Non-persistant groups: = counts, plugins
• Store transients in database is also checked.

Browser Cache settings

• This has a lot of combos and not sure what is custom and default. but I do see expires header and lifetime of 3600 here too.

If I had to take some guesses

• makes me feel it is an expires things?
• physical memory + redis arent configured (?)
• not grab everything (ignore woo)
• Conflict ( I use Avada Theme, & Jetpack but I’ve made sure to turn off conflicting settings like gzip. I have search console access and had to install redis myself and maybe configured it incorrectly as I was originally using the ‘Redis Object Cache’ plugin by Till Kruss )

REDIS STATS INFO

I have checked $redis-cli info stats and these are the values its provided

Total_connections_received: 821

Total_commands_processed 361million

total_net_input_bytes 890billion

total_net_output_bytes 3Trillion

expired_keys: 61million

expired_state_perc 16.31% | | Evicted Keys: 407k |

expired_time_cap_reached_count 1,027

keyspace_hits: 124million || keyspace_misses: 126million

I’ve called $redis-cli monitor and it always shows a lot, and fast. no time to really read but i can see entries that include W3TC in it.

Any and all help is useful (: I’m willing to read documentation and provide more info / status reports. or if this is too personal for these forums maybe a more direct channel to contact W3?

I have purged the website cache using W3 Total Cache. Since I use Cloudflare, I “purged everything” there as well. However, the error is persisting. I am reactivating the plugin just so my site doesn’t stay offline, but can you provide steps that will help me figure out why this is happening and how I can remove Autoptimize?

See

Introduce wp_cache_get_multiple() in WordPress Core

Caching improvements in WordPress 6.0

Performance Field Guide for WordPress 6.1

These changes should be reflected in this plugin,

It’s maybe something i can change in the W3 Total Cache, but for now this info could be useful. Don’t have much time.