OGS connection issues for BT users (Was: Is OGS running slow? ... no, it isn't)

thats - it. You can see the timeout doubling from 2000 to 4000 to 8000.
Not sure if that is in ms - as 8000 would be 8 seconds. Pings are only taking 17ms.

What is happening with those messages is that we are trying to send a message FROM your browser TO the OGS server.

Normally the timeout for that is milliseconds. When we start having problems we increase the timeout, in case things just went slow for a bit. By doubling it. At this point we put up the sign saying ā€œSlow internetā€.

When it is ridiculous (10s or something like that) we give up and say ā€œConnection lostā€.

Hopefully some folk who know how to interpret the results you’ve shared will be along soon.

1 Like

This seems to add to the feeling that the problem is in your internet connectivity.

The forums come from a completely different server than OGS.

2 Likes

@hughb since you have a chromebook there can you see if you can run a traceroute from the shell and paste the results?

it should be something like

mtr online-go.com

or

tracepath online-go.com

or

traceroute online-go.com

Also, can you confirm that when you are experiencing the slow/unresponsive behavior on your phone that you are away from your house or otherwise have wifi turned off? (That is to say, that your phone isn’t actually using your home internet connection or another BT link?)

1 Like

Server, data center, and provider to be clear. The only thing in common is the use of Cloudflare, so I’m suspecting that if folks are not seeing slowdowns with other sites but are seeing slowdowns with both the forums and the main site then there might be a peering issue between their ISP and cloudflare, it might be worth bringing it up with the ISP.

2 Likes

Thanks for that but when on Pandanet for example it works a treat and on other forums I use. I have checked my connection and speed i the same as ever and there have been no BT issues recently . My TV streaming work fine as well. None of the problems with the OGS forums or site happened untill a couple of weeks ago

@go-go-ghost I think the test would be to check other cloudflare sites at this point. Pandanet uses Akamai, and your TV streaming networks are going to have their own systems and peering arrangements with your ISP.

Can you visit here: https://speed.cloudflare.com/ and see what you get? (Anyone experiencing issues really, same for the traceroute info if you have that ability)

For all six online-go.com IP addresses I get remarkably and consistently fast tracert times, all the way to the final destination.

Furthermore, when simply displaying a game, the duration and latency times for all HTTP requests are ~300ms or less. So the network and site are performing really well, except when I submit a move, which sounds like a database write issue.

2 Likes

Have you shared a screenshot - I didn’t see it?

Console log and network timing would be great, and will tell the difference between network and db issues.

most of the handful of people reporting this problem are in the UK

Same here – in the UK with BT

1 Like

I can see how you come to that conclusion but it’s just not the case, we’re processing on average about 2250 moves a minute just fine right now. It’s not on the OGS servers, this is very specific to a small subset of users in the UK it’d seem.

The other variable is that we use websockets for all game stuff, and sometimes ISPs try to do stupid things with websocket connections - although it’s been much less of an issue in recent years than it was a decade ago. When they muck around with websockets, the web requests can load fine but the websocket connection can be unstable and drop frequently. It sounds like the site is loading slowly all together though which is why I’m suspecting there’s something going on between BT and Cloudflare, or potentially a leg of the internal network in cloudflare somewhere in the UK, but I’m not seeing anything on the cloudflare dashboard.

Could you run the https://speed.cloudflare.com/ test @denishowe?

1 Like

https://speed.cloudflare.com/ says
Down: 50.1 Mbps
Up: 9.1 Mbps
Latency: 7.00 ms, down: 17.0 ms, up: 453 ms
Jitter: 2.74 ms, down: 4.00 ms, up 51.3 ms
Packet Loss: 0%
Server location: London
Your network: BT (AS2856)
Your IP address: 2a00:23c7:f7bd:7e01:35e4:9345:f745:ffd3
Though I’m not sure any of that tests websocket performance.

1 Like

Interesting, that up ms of 453 ms is really high, everything else seems good though. However, if there’s something going on sending data, well that kinda lines up with problems sending things like move data, potentially.

I would expect both down and up latencies to be pretty close, it’s expected to have a bit worse up latency since links are asymmetric, but no where near half a second to send something, that’s really high. Mine for instance is 22.5ms down and 30ms up.

Not sure if you pay for bandwidth or not, but if you wanted to you could use a different speed test system to see if the results are consistently bad or if it’s a cloudflare specific thing, for example https://www.speedtest.net/ will probably serve out of a non cloudflare location and will provide your latencies as well.

2 Likes

https://websocketking.com/ is returning websocket replies in <10 ms.

2 Likes

Yeah see that 31ms upstream latency? That’s more like what I’d expect.

Thanks for the websocket test too, that’d seem to imply BT isn’t doing anything funky with websockets.

So the prime suspect that I see is that upstream latency between BT and cloudflare. That should be more like the 31ms you’re seeing on the other speed test. It’d line up well with basically being able to browse just fine pretty much the whole internet, but when you send data to any cloudflare backed sites specifically, such as submitting moves to OGS, that’s when you’d experience the issues.

You might try contacting BT about it, the fact that it’s been going on for a few weeks now is a bit suspect, they might not be fully aware of the issue since the downstream from cloudflare is working great, and most folks aren’t going to notice upstream issues beyond some increased request latency.

3 Likes

I’m no expert but when we were with BT ā€œfibreā€ broadband (FTTC, missold as FTTP) we always had an apparently fast connection but suffered frequent random drop outs/weird internet behaviour. BT ended up , one bit at a time, replacing various components from cabinet to the house. The only thing that resolved it was when we switched to LIT with a proper FTTP connection!

1 Like

I suspect the link from the home to BT is ok in this case since the other speed tests are consistent and fine. (Also, multiple BT customers are experiencing the same issue it’d seem).

I have now tried my laptop, phone and kindle at a friends house who also has bt internet and exactly the same problems so it appears not to be my conectivity

Pretty sure the problem is an upload issue between BT and Cloudflare @go-go-ghost , see the above with @denishowe

I hope you’ll forgive me for torturing your servers but, just for fun, I tried connecting from websocketking.com to wss://ai.online-go.com and sending a few
["net/ping",0]
I received responses for 20s after connecting, then nothing, then, after another 20s, it disconnected.
This was entirely repeatable.
If I left the connection idle, it disconnected 30s after connecting.
The result was the same for online-go.com.
My route to websocketking.com doesn’t involve Cloudflare (it seems to be on AWS) but obviously, I’ve no idea how they connect to you.