Looks like the WACS route is flapping - upstreams are on it, I think they're going to dampen until resolved.
-
Join the MyBroadband community
South Africa’s biggest forum. Discuss, discover, and connect with thousands of members.
-
Question of the Day: Did you buy your last car with cash or on a finance plan?
Web Squad ISP
- Thread starter Gazed416
- Start date
- Status
Looks like the WACS route is flapping - upstreams are on it, I think they're going to dampen until resolved.
GoneWithTheWindows
Honorary Master
Is this resolved?
Still seeing strange routing, mybroadband for instance is fine 20ms, but carbonite is 165ms.
Also international feeling a little slow.
I think Cloudflare turned down some prefixes and services in SA today. Saw reports of 1.1.1.1 jumping to JHB at some point. Carbonite traces through cloudflare's transit via CPT. Can't really comment on Cloudflare's routing decisions - but two flaps on WACS (their primary route) in one day, would maybe make them turn down some services until they're sure things are stable
With regards to transit, upstreams are monitoring the WACS circuit and there is some work going on in the background. Will chat to them now to get a better feel for their progress.
GoneWithTheWindows
Honorary Master
Is this kind of speedtest "ok" on a 1Gbps line?
Looney
Executive Member
Definitely not
Try a server we don't pick up through peering (IE, SA ISPs' own servers in the UK) - this is more a sign of the peer's transit than ours (the peer carries this traffic to and from London and hands over to us on NAP). Our London speedtest server is hosted on Digital Ocean (a reputable cloud provider with a pretty good network). You can use our test (speedtest.lon.websquad.co.za) -we use this to monitor if our upstreams are having peering issues in London. Or Speedtest.net coreix is pretty reliable.
Edit: Could also be packet loss on your line - i see you are on GPON, so we can't test to the Vuma handover and a low ms test won't make much of a difference anyway. Test a Cape Town and a Durban server (download the speedtest.net app) and see if there's any significant degradation in performance there.
Edit 2: Picking up on some upstream issues - will confirm further. But above for other ISPs international servers still applies.
Last edited:
Looney
Executive Member
Definitely something wrong with international. I’m actually starting to get sick and tired of this. I’m paying for a 1gb line and I never get 1gb speeds.
Definitely something wrong with international. I’m actually starting to get sick and tired of this. I’m paying for a 1gb line and I never get 1gb speeds.
We're escalating to our upstreams - also seeing a drop over the last few mins. Some test destinations are good, but others not so much.
@websquadza
What is the id for speedtest.lon.websquad.co.za as it is not listed on `speedtest-cli --list`
It isn't setup as a Ookla server. HTML and iperf only - Ookla works better in a managed environment, not a cloud.
Last edited:
International Transit Notification:
One of our upstream providers has picked up on an issue and will be conducting Emergency maintenance on their WACS path tomorrow:
Date: 17 September 2020
Time: SAST 09:00 - 12:00
We will steer traffic via our other transit providers during this time - there remains a risk of elevated latency as some traffic may route via the East Coast.
One of our upstream providers has picked up on an issue and will be conducting Emergency maintenance on their WACS path tomorrow:
Date: 17 September 2020
Time: SAST 09:00 - 12:00
We will steer traffic via our other transit providers during this time - there remains a risk of elevated latency as some traffic may route via the East Coast.
Transit Update: WACS have started maintenance on the affected ports. Please expect elevated latencies as some routes are moved around.
Damn you're not kidding. I didn't realise "elevated" meant +100ms
Yeah, we've brought on some additional routes to compensate for the downtime - however, the East coast paths are a) longer routes (Both physically and that they land in Marseilles, then onwards to London), and b) you're in CPT and there still aren't any direct routes to Mtunzini from CPT, so that adds a fair bit of latency while traffic crosses to JHB > DBN then onwards.
GoneWithTheWindows
Honorary Master
@websquadza Over the past couple weeks i have been having my net drop every now and then I couldnt pin point it though as things like my work resources and voip phone would drop but by the time I got MTR up it was back, caught it now while doing some MTRs for your support desk -
Will look into this. Apologies, confused you for having GPON yesterday, you are on the Vuma trenched network. That highlighted hop is a Vumatel router that doesn't always respond to ICMP - it's not causing packet loss. What's interesting is that you have some PL carrying through. We're running tests to the above IPs from our side to isolate a cause.
GoneWithTheWindows
Honorary Master
Just curious how does a device not always respond to ICMP? Its either open or not right?
Generally ICMP is a pretty low priority protocol, and some core devices have rate limiters on them for ICMP - so only x number of requests per second are allowed and the rest are dropped. A core device on Vumatel's network is more likely to receive a flood of ICMP requests and as such, a rate limiter is pretty good idea - you want this router forwarding traffic, not spending most of it time replying to pings. A MTR tests is basically a tool that sets individual ICMP sessions to each router it sees along the way, so it doesn't actually track the packet as it moves through each router and report on the full path - if that makes sense. So think of it as more of a overview of what is happening and not an actual reflection. But it's pretty useful for troubleshooting obvious issues.
GoneWithTheWindows
Honorary Master
Awesome, thanks for that explanation
@websquadza Any update on the ETA of resumption of the normal routes?
- Status