Issues with Perth Pop

Website, Application, Performance DNS & Network
1

Am posting this as an end user, because i am having an odd issue.

The issue can easily been seen with the game Eve-Online.
While playing the game, the client will disconnect at seemingly random times.
Then within 30mins to 3 hours, it will happen a second time.

I noticed that during that 30mins to 3 hours, my route to the Eve Server changes and my ping increases from ~7ms to 50+ms.

This is my normal Trace Route

Tracing route to d638e439e07f413c97200d057e5ebd05.pacloudflare. com [172.65.201.188]
over a maximum of 30 hops:

1 <1 ms <1 ms <1 ms 192.168.1.254
2 9 ms 8 ms 11 ms loop612451320.bng1.vdc01.per.aussiebb. net [61.245.132.1]
3 7 ms 8 ms 7 ms hundredgige0-0-0-7.core2.vdc01.per.aussiebb. net [202.142.143.44]
4 7 ms 7 ms 12 ms as13335.wa.ix.asn. au [198.32.212.107]
5 7 ms 6 ms 7 ms 172.65.201.188

Trace complete.

This is the Trace Route when my ping is high

Tracing route to d638e439e07f413c97200d057e5ebd05.pacloudflare. com [172.65.201.188]
over a maximum of 30 hops:

1 <1 ms 1 ms 1 ms 192.168.1.254
2 10 ms 5 ms 7 ms loop612451320.bng1.vdc01.per.aussiebb. net [61.245.132.1]
3 55 ms 54 ms 52 ms hundredgige0-0-0-7.core1.vdc01.per.aussiebb. net [202.142.143.46]
4 52 ms 51 ms 54 ms be2.core1.nextdc-p1.per.aussiebb. net [202.142.143.151]
5 53 ms 52 ms 54 ms hundredgige0-0-0-19.core1.nextdc-s1.syd.aussiebb. net [202.142.143.54]
6 52 ms 52 ms 55 ms be1.core2.nextdc-s1.syd.aussiebb. net [180.150.1.117]
7 57 ms 59 ms 53 ms as13335.syd.edgeix.net. au [202.77.88.7]
8 52 ms 51 ms 52 ms 172.65.201.188

I have contacted my ISP (Aussie Broadband) who cannot see an issue on their end.
I have verified the issue with another user who is with a different ISP, so i don’t believe it is an issue with my setup or with my ISP.

As an example, it happened again today between ~13:30 - 15:15 +08:00 UTC 14/07/21.
There doesn’t appear to be a pattern to it that might indicate scheduled maintenance.

Obviously i have contacted Eve Online, they’ve come back with the usual “make sure you’re using ethernet cable and not wireless” and are even advising me to use a VPN to get around this issue.

While i acknowledge that someone playing spaceships over the internet likely isn’t a huge priority, i imagine other users are having issues too and due to the intermittent nature of the issue aren’t sure what is going on or what is causing it.
If it was predictable or less often i wouldn’t be too worried, but it has happened 3 times in less than a week.

2

And they did not notice the third hop already, where it jumps from ~5 milliseconds to a tenfold?

I am afraid that’s very much an issue with your ISP and they need to look into this.

1 Like
3

So you’re saying it’s normal for the cloudflare pop to randomly go offline multiple times a week for hours and have traffic routed to another city?
Unfortunately at this time i don’t have a tracert from a different ISP when it is down, but as i said they see the same behavior.

4

What I am saying is that whatever issue you are experiencing is something your ISP needs to fix, based on the routing information you provided.

5

But why is the route changing so often? I wouldn’t think that is normal.

and again, why does it happen for other ISP’s if it is my ISP’s problem?

6

That’s a question for your ISP, but - as evident from the available information - the issue is not even that the route changes but that you have the increase in latency already within your own ISP’s network.

7

The latency is due to communicating between 2 cities that are almost ~3850km apart (thats almost 2400 miles).

As you can see here, that latency is about what you’d expect

Ping time between Perth and Sydney - WonderNetwork

You’ll notice a similar ping for other cities that are a similar distance away, say New York and LA, Barcelona and Moscow, etc.

again, my problem is not with the ping itself but the fact that the route keeps changing. If the route didn’t change there would be no issue.
If it was my ISP’s problem, it would not happen for people on other ISPs that have separate networks/infrastructure.

What would cause it, is an endpoint for a particular service (say, Cloudfront) going offline and then me needing to route to another endpoint on the other side of the country.

The next time it happens and i am not at work or out of the house, i will endeavor to get trace routes from other people with other ISPs.

8

Once more, that’s a routing issue internal to your ISP and can be only fixed by them. Forward the traceroute once more to them and point out what I wrote earlier.

9

So, multiple ISPs all somehow have the same routing issue at the exact same time?

10

I am not commenting on other ISPs, I am commenting on what you provided so far.

11

well i am.

As i have said, it happens at the exact same time with someone else on another ISP.
That shows me it is not an issue with my home setup, my internet connection or my ISPs network.
The increase in ping is due to my traffic being routed to a city on the other side of the country. I’ve shown that ping is about right for where the traffic is going.

Ok, for a second, forget all this.
What would happen to this traffic if the Cloudflare node/pop/whatever in my city goes offline for whatever reason?

12

Well, then you would need to provide information about these connections. In the case you originally mentioned - and that’s what we’ve been discussing all along - the issue was clearly with your ISP and not Cloudflare.

As for your other question, when a PoP goes offline it will be marked as re-routed and the IP announcements from the next best location should take effect.

13

In my first post i did say

I have verified the issue with another user who is with a different ISP, so i don’t believe it is an issue with my setup or with my ISP.

I will get that additional information when i can. It happened both days of the weekend last week, so hoping it will happen again. I am monitoring it using Telegraf, will see if i can set up an alert to notify me when it happens.

As for your other question, when a PoP goes offline it will be marked as re-routed and the IP announcements from the next best location should take effect.

Could that look like what i am seeing? The route changes to go to another city (in this case Sydney) which is believed to be the next best PoP?

14

I don’t think so, because the latency goes up to 55 milliseconds within your provider’s network and then stays there. In the first case you had seven milliseconds, in the second 55. That’s the thing your ISP needs to clarify.

If that hadn’t happened already within your ISP’s network, the routing could have been the case but that’s something that’s covered by Deprecated - Peering - Why don't I reach the closest datacenter to me?.

But again, I do not think this will apply in the mentioned case for just mentioned reason.

15

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.