NGINX At Scale

twest

I've reread a bunch of threads about the woes for nginx users, seems some critical functionality remains out of reach to get the maximum benefit, like being able to customize the config.

I'm curious though, is anyone using it at scale on the enhance defaults? I'm curious about both mass hosting for small sites + server density, and also for hosting large/busy sites. Is anyone successfully hosting at scale with nginx here?

I'm thinking if I want to roll out some nginx servers or throw down the $ for litespeed 😃 It would be great to save the expensive cost of LS if nginx can handle high load (like one WP site with 100k daily sessions!). I do love me some lsphp on cloudlinux servers tho, so litespeed is tempting!

JohnB

twest I'm thinking if I want to roll out some nginx servers or throw down the $ for litespeed 😃 It would be great to save the expensive cost of LS if nginx can handle high load (like one WP site with 100k daily sessions!). I do love me some lsphp on cloudlinux servers tho, so litespeed is tempting!

Nginx can be configured to be faster than litespeed and supports micro and full page caching BUT... even if enhance gave access to all parts of the config...

There are issues where plugins and themes sometimes don't work. For example a common image optimizer plugin for webp requires an extra step for nginx. The average user won't like this I imagine.
No htaccess which many plugins rely on. Plus side: a lot of malware relies on htaccess
many caching plugins will not play nice with nginx caching or varnish if you went that route.

Personally I host what I build and manage it so these are not an issue. If you are offering mass hosting, I would consider the above issues.

charlesalmeida

Hello, I have an opinion with an artists' website that has many sessions. NGINX only helps with Full Page cache. Enhance lacks some essential settings, such as setting cache time, browser caching. Not to mention how essential it is to have a cache with a local domain purge hook via plugin.
And worry about setting ssl with high security score.

DracoBlue

Unfortunately, the enhance panel for the Nginx server lacks many crucial features needed to run it in a production environment. Configuring most scripts is not possible unless specifically tailored for WordPress. I don't know when Nginx management will become more comprehensive, but regrettably, I'm forced to use Apache.

DracoBlue

Today, I migrated the server from Apache to OpenLiteSpeed. In my opinion, OLS is more efficient and faster than Apache, especially when combined with the LS Cache plugin for WordPress. I consider it a great alternative to, for example, NGINX. I would be willing to move my server to NGINX, but currently, it's not possible because Enhance provides too few configuration settings.

twest

DracoBlue ah damn, yeah that's what I was thinking about nginx.

I really like the litespeed php handler, I use it on cpanel very successfully. I couldn't use OLS yet since the configuration page doesn't "stick" between updates. For example, I don't think 1 worker is sufficient to handle my servers volume of traffic, it would crash. I guess it wouldn't be terrible to run updates late at night so that config can be updated right after, but that's a lot of extra manual work per server - it would make more sense to just get enterprise litespeed so the config persists.

I've actually been looking at some litespeed licenses to test out. I like the license available at cPanel, $45 for 3-workers, that might just do the trick for my big servers.

But alas, I have a very well optimized Apache config, so I'm not sure if I'd really see any difference to make it worth it. I am sure it would lower general server load though, as lsphp is so much more efficient, so maybe it would be worth it just to reduce wear on my servers.

twest

JohnB all good points. Considering Enhances integration status, and compatibility with things many users need, I'll stick with Apache and Apache derivatives.

I'm actually heavily leaning on going with some litespeed servers at the moment. As a drop-in replacement for Apache at least there won't be a bunch of issues to fix in existing configs. My only issue is compatibility problems with WP rocket (it needs an extra helper plugin installed, or else pages won't load for visitors, instead a blank file downloads), and a lot of my clients use wpr, so conversion would be a pain either way (add the helper plugin or convert them to the lscache plugin).

I've been toying with the idea of converting Apache servers to litespeed, and then turning off the server side caching in ls config to prevent WP rocket conflicts... Of course that gives up the efficiency of server side cache, BUT it allows for use of litespeed's PHP handler, so there'd still be a big improvement in efficiency and speed.

JohnB

twest BUT it allows for use of litespeed's PHP handler, so there'd still be a big improvement in efficiency and speed.

That might work. The main positive and less problems. Personally I hate the lscache plugin even though the performance is really good. Its the annoying bugs like how sometimes the cache for the homepage just doesn't clear. It says it cleared but it doesn't no matter how many times you click it. You have to delete the files manually (their support advice). This can be problematic with auto updates.

And they are dishonest in their benchmarks by not running the same hardware or configuration differences but that's marketing I guess. The garbage that is cyberpanel is implicitly recommended because its on their website.

All that said the performance is great. I wish they would clean up a bit so its easier decision lol.

twest

JohnB I went ahead and plopped litespeed on a server last night. It turns out the default config for enterprise litespeed is pretty great! The server side cache is disabled by default it seems, because you have to specify a directory for the cache in the ls config panel.

I only had two sites on that server, but both switched over good, no WP rocket conflicts.

I will convert a large server tonight and see how performance and load go 🙂

Rich

JohnB Strange I've never had that issue before and I've been using litespeed + LSCahce for at least past 7 years. There was the issue of it not purging the cache properly after WordPress ToolKit on Plesk did updates, but there was helper plugin someone wrote that sort of resolve that.

@twest I've had litespeed enterprise (default config) easily handle 250K hits in a day on a single site, server load barely even rose. (this was 5 years ago) That was with a single worker process license on a dedicated server, with CF Pro serving Edge cache. E1270-v3 was the CPU if I remember correctly.

As for using LSWS or OLS, you're likely leaving performance on the table by using caching plugins that aren't LSCache with litespeed. As you won't get server side cache options.

As for lsphp, I too thought it was awesome and loved cloudlinx + lsws. But ehance separates out the PHP layer into individual containers (customers) I don't think it will be using lsphp, as it's using php.fpm instead, the webserver sits atop of that. https://enhance.com/docs/php/ (someone correct me if I am wrong)

I've been running OLS for nearly a year now on enhance with some small and large sites. No issues to date, except Wordfence auto-update not working due to .htaccess reloads. (now implemented in recent appcd update)

twest

Rich nah when you switch to litespeed you also switch to lsphp for php handler. Though I'm not sure exactly how enhance implements it, I've never had to customize lsphp config since it scales up on its own. I think it's max php connections per user is limited to 10 by default though, which would need to be increased for busy sites.

The bug specifically has to do with how WP rocket implements _gzip in the htaccess, or at least that's the most common culprit. The WP rocket helper plugin for litespeed servers clears the htaccess of all WP rocket rules. Someone mentioned disabling litespeed server side cache also prevents the bug from happening, so that would be preferable for me in the short term until I can get all the sites switched over to litespeed plugin or add the helper plugin.

I only ran into the bug when I was testing the different webservers and tried out OLS. I presumed the problem would also be present in enterprise litespeed. I wonder if enhance enables the server side cache in default OLS config, but not for enterprise? That might explain why I wasn't affected in my changeover last night. Or maybe it's just totally random, idk, I just know it affected me during OLS tests lol

Rich

twest Nice ok, thanks for sharing that about the lsphp handler for enhance that good to know. Honestly I've not gone looking/tweaking OLS config knowing it's volatile right now.

LSWS/OLS uses brotili rather than gzip, might that have any further bearing on that bug?

Good luck with the tests 🙂

twest

Rich I'm not really sure what the specifics of the bug are about, I've read about it as a known issue going back +4 years, and it still exists lol

When you switch to enterprise LS, enhance gives you a handy password change tool right inside the enhance UI with a link to the litespeed config page on port 7080. They should have that for OLS too, but I guess they don't think it would be useful since the config would get wiped during any vhost update.

I actually really like the ls config area, so many options to find tune things. To me it's easier to scroll through options in a UI rather than reading config files and cross-referencing the manual for every little thing. Default config files usually don't have all the options in them either, so you gotta manually paste em in. With LS it's nice because everything is there in the config panel, there's a helper popup to describe every option and it's all logically laid out in sections.

I saw the brotli integration that's on by default, it also has gzip too, so I'm extra puzzled about that bug... Actually I recall someone mentioning it could be that Rocket adds it as _gzip and LS uses -gzip? Whatever, it's a silly bug and easy enough to work around for now 😃

Rich

+1 to LS Console. Also, I chuckled at

They should have that for OLS too, but I guess they don't think it would be useful since the config would get wiped during any vhost update.

Yeah, I am sure it will come in time, parsing the OLS conf is a pain I've written my own Auto WP site generator OLS for local development years ago. Always got double white spaces in the config on update/save stuff like that. it's plain text, not xml or other type of structured file. So harder to handle.

Here's to hoping though, as I do sorely miss that capability. Sure tweaking the webserver, but it's the reverse proxy I miss. Loved spinning up a docker compose for some new project.

Speaking of which, since we're discussing nginx thread, I did hunt about and see once a very clean separation of vhost files, that made an ideal target for cron to replace it push come shove. I planned on patching a single node with nginx to support docker reverse proxies. Sadly not found the time yet.