Yeah, there is a lot of connectivity in Holland and some fairly major data centres. So, it’s probably down to the capacity of the mirror sites themselves rather than the connectivity.
I use Digital Ocean for some hosting. Not all services are available in London, for some things I have to put stuff in their Amsterdam DC.
London isn’t necessarily as “big” in the World as sometimes it makes out … 
Looking at it now, it’s actually quite bad news for Data Centres, so this is going from an instance on a rack or blade server that will almost certainly have cost thousands and be stuck on some relatively expensive bandwidth, to a £78 RPi sat on relatively inexpensive broadband;
AND, Sunday is generally the busiest day in terms of random traffic for the Forum. Google ranking is affected by response time, wonder if this will make any difference … although I have also to guess that the difference is partly 4 CPU cores so any monitoring request if 4x as likely to get a free CPU core for max performance. To think I was worried about not running it on an AMD based server … (!)
I added “Bookmarks” and “Hot” to the navigation options at the top of the page and re-ordered for aesthetic effect.
I’d not picked up on Bookmarks previously, but there is a bookmark button at the foot of every thread, once you bookmark a thread, it’ll appear under the “Bookmarks” tab.
Forehead slap time.
I just started seeing 429’s from the Forums (too many requests). Looks like it’s wasn’t seeing the incoming IP properly so all requests were effectively coming from the same address (cloudflare) and running into the rate limiter.
I have now tweaked.
From my previous stress test;
- Live copy on Digital Ocean, 78 requests per second
- Desktop copy on RPi5, 74 requests per second
Having now tweaked the rate limiting, I’m seeing 144 requests per second (!)
Is it possible to tell which of these request are crawlers or genuine seekers?
144/sec seems quite high to me.
Sorry, 144/sec is my benchmark, that’s not the server load.
I’m guessing moderate usage from a few dozen people is enough to trigger the rate limiter if all requests seem to come from the same address. I suspect (!) that the rate limited was cutting in quite a bit which many have slowed down general load times.
Seems a little quicker now for me …
I’ve been trying to see what was “wrong” with the CloudFlare / rate-limiter config, but it all looks Ok as far as I can see … still, tweaks seem good, will re-visit when I have more time.
Ok, so it all seems to be working Ok, response times look Ok and we’ve not had any outages since the switchover …
So … couple of days worth of pain, however … the upshot is that I’ve moved the linuxforums connection from a Cloudflare Zero Trust tunnel, over to my own tunnelling infrastructure.
If anyone sees any problems or perceived performance issues please let me know, but thus far it seems marginally more responsive.
All of a sudden (!) Cloudflare tunnels seem to have developed an issue with WebSockets, and although this didn’t directly affect the forums, it’s a breakage that’s wraught havoc elsewhere and I’m now wary of other issues CF tunnels my carry. (and I’m not sure whether any of the Forum plugins use websockets)
The websockets applications I’ve moved not only now work properly, but they seem a lot more responsive.
Use the cloud they said, it will take all of the work out of it they said, it’s much easier they said, you can just set it up and it will work they said. In Cyberspace, they can hear you scream.
