Skip to main content


 

Connections seem a lot slower the past day or so


!iSurf Support Is it just me, or is anyone else finding connections to isurf.ca slower than normal?

I know the nature of the internet is that so many things vary at times, but I'm suddenly seeing it a lot slower anytime I connect. I'm not experiencing the same kinds of slowdowns on other sites I frequent.

I really wonder about DigitalOcean's setup in Toronto. From where I am, the connections are normally excellent, but I get occasional reports that things are pretty disastrously bad from others like yourself at random times.

It's been just fine for me the past couple of days when I've been using it, but please let me know if it continues to be an issue.

It seems OK right now. Next time it happens I'll run an mtr test to see how the path looks.

This what it looks like right now;
Host Loss% Snt Last Avg Best Wrst StDev
1. router.asus.com y mode 0.0% 57 0.3 0.4 0.3 0.5 0.0
2. lo0-0-lns04-tor.teksavvy.com f 0.0% 57 6.2 7.7 4.6 46.9 8.0
3. ae0-2150-bdr01-tor.teksavvy.com 0.0% 57 5.4 8.3 4.8 51.2 7.5
4. digitalocean.ip4.torontointernetxchange.net 0.0% 57 5.2 5.6 4.7 10.2 1.0
5. ???
6. social1.isurf.ca 0.0% 57 5.5 5.4 4.9 6.2 0.2

Tough to get better than that... So I'm not sure what's caused the issues for you this evening, and I haven't been able to replicate it and I'm at 60+ms from the server.

just encountered it again.

No real change in the mtr report, still very good ping times and no loss. I'm wondering if the server end (web server or authentication, just a guess though) is taking a longer time to respond when connecting. I'll have to setup wireshark and see if it shows any insight into what may be happening.

I don't know what you run for collecting real-time metrics on your end, but I highly recommend netdata. I proxy it behind nginx so I can password protect it and run TLS. It is surprisingly lightweight for the data it collects and gives a lot of useful information even with defaults. Fine tuning the configuration(s) is usually pretty straight forward as well, with lost of existing collectors available.

https://my-netdata.io/

I tend to configure netdata to collect and keep a running window of 25 hours. I figure that is a long enough window to be able to review to see what may have been happening. For that netdata uses approx 500MB of RAM. Pairing it with Graphana allows seeing metrics over a longer period of time too if you want.

I ran netdata on the server in question for several months, pulled it off when I was getting RAM constraints (which will be resolved in the future with a size upgrade to the server, of course). Everything looked really good, but yes, I'm missing out of some of that real-time data currently.

Page load time 36.2 seconds this morning.

I'll start running some dev tools to see if I can monitor each and every item being loaded. I suspect a major contributor may be one or more external links that is taking a long time to respond and holding things up. If I see any pattern or anything that jumps out I'll let you know.

Hmm, well, friendica does do caching and proxy caching, but I think it has some limitations as to what it can cache or it can recognize as cacheable... and I forget whether that is based on the type of embedding or if I'm just mis-remembering.

That said, perhaps I could see a reference to a post that has something from a slow-loading server, and I can check in to whether or not friendica is making a local copy, and if so, when. (ie, would be nice to determine if friendica is doing the appropriate thing with helping speed things up, which it sounds like it may not be)