New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Ever heard of RIPE Atlas?
Again, this may fly with your uninformed customers. Here everyone knows that P2P piracy is not a big deal.
In my humble opinion, this thread is reflecting badly on you. I do not have anything against you and at some point in my life I would've been a potential customer, but just leaving this thread as is and not dragging yourself even more into this hole could prove to be a smart move.
What is the PSA about? That your upstream has issue with Telia, potentially due to saturation?
That’s not what you say here.
@PulsedMedia
You bash and slander them, even in the title, yet you fail to provide conclusive evidence. You can't due to e.g. privacy concern? No problem, but then don't start a thread like this.
@Clouvider is a known and respected entity around here and AFAIK has a total capacity of multi Terabits per second. About your capacity I do not even have any information other than some of your nodes/dedi seem to have dual port 10 Gb/s, which is nice but doesn't provide any clues as to your backend/uplink capacity.
That's not saying that you lie; maybe there are some local network problems, I don't know. But I do know that it's not acceptable to bash/slander a very well reputed provider without clear and proper evidence and I think that a well known provider with Tb/s connectivity saying they have no significant problems with Twelve99 weighs more than wild allegations without proper proof.
well below congestion metrics, typically we have around 50% daily average, and weekly burst (5min average) does hit about 70% in the past month.
Admittedly, this has been badly worded. We should have began with that this specific hop, not overall Twelve99. Albeit, we have received reports of very odd and peculiar routings elsewhere too; But do not have access to neither end networks to double test these peculiarities.
TCIC is working on it finally, sadly only after the magic word "lawyer" however. They ignored 100% of all previous reports with MTR both directions, with all of the IPs etc. Sorry we are not comfortable of releasing customer IPs publicly, only for the parties working on the actual issue.
Regardless, this is a real issue for all of the accessibility in Finland for Twelve99 / Telia. It's been getting slightly better, but personally i suspect that is because so much was moved away from Twelve99.
Really tought i was doing the community a service telling this issue exists
Who is “we”? What about your Upstream to Telia? As this is where the congestion looks to be from the limited information available? Surely you are not privy to this information?
Looks like the problem is between FNE AS47605 and Twelve99 (without seeing the path back)
if they are not congested then maybe a faulty optic between both or some issue on the fiber path.
Moved from https://lowendtalk.com/discussion/comment/3365913/#Comment_3365913 thread.
This is a complaint against Twelve99, not a complaint against PulsedMedia.
Twelve99 is doing well in dropping 40% of the packets.
Setting
-u -b 50M
or-u -b 10M
or-u -b 1M
results in the same 40% packet loss.Sad, but even having BBR wouldn't save from 40% packet loss.
FNE <> Twelve99 is far from congested, has many 100G links, with lots of headroom and no errors. ... and this is just the hls-b1 link, afaik FNE has another one at another location. We are at the same building where the hls-b1 link is (and one of the reasons we are in that building! ALL the sea cable fiber from Stockholm terminates to that building)
Yesterday Twelve99 started doing something, routes were "flickering", going off and then coming back. We've been rerouting all known problematic networks off, but yesterday the Twelve99 link was disabled completely as they were working on something and triggering many alarms.
Guess what happened?
second highest 24hr peak, and by far highest 24hr 95th % utilization on our.
Regardless, the issue is and has been with Twelve99 beyond shadow of doubt.
Yikes! That's huge and even worse than usual! It's usually 6-7% loss. Please open a ticket for both distro upgrade + getting you rerouted to Level3.
hls-b1 worked for a moment, then not again. Disabled that link, so rerouting through another telia router elsewhere, and bandwidth utilization jumped up again. Record level throughput upstream.
"slight" jump after disabling Twelve99 hls-b1
How can you disable a link?
The only provider of AS203003 is AS47605, who can decide where egress traffic should go.
Or, do you control AS47605 too?
Egress to Verizon FiOS is now using AS3356, which works perfectly.
I provided traceroute of both routes in Ticket 212018.
as said, we have good relationship with AS47605. Not much traffic went through the hls-b1 peer overall for them, but all of our Twelve99 did as we are on the same building - shortest route and all that
So they disabled it, and rerouted through their another link elsewhere with Twelve99 and all of sudden things work.
I'm seeing Level3 being selected for FiOS and most of my servers in USA.
VirMach Los Angeles is single home Twelve99.
It's using hls-b4-link.
<1% loss on UDP; 146 Mbps on TCP.
So, the problem is limited to hls-b1-link, then you can't accuse Twelve99 for having huge congestion all over the network.
It could be just a single router or line card.
It could be just a single router or line card.
Never meant to "accuse their whole network is congested". Huge network, it can have issues on small segments, does not mean everything is not working correctly.
@yoursunny can you provide an mtr both ways?
This shows no packet loss though. If one of the paths has an issue then both mtr should show loss on the final hop above 0.
Perhaps you could do it on a higher count, like a 100 or 1000?
There's no loss because hls-b1 has been disabled.
Not much point to go through the mtrs then. I’m very interested to see one when the condition is ongoing.
Thank you anyway :-)
Routing to FiOS is on Twelve99 again.
The dreaded hls-b1-link got enabled again on one direction.
mtr
shows no packet loss; sending more packets (-c 1000
) has same result.Note: FiOS blocks outbound ICMP traceroute, so that I can only UDP.
TCP performs badly.
The particular server is Debian 8, so there's no BBR.
Submit ticket with your IP and we will reroute you to globalconnect so we can redo these tests.
It's Level3 again, and TCP is smooth.
@Clouvider , what's your verdict on the "mtr both ways" going through hls-b1-link posted above?
No packet loss but TCP has so many retries.
Can we blame Twelve99?
Thanks for the metrics after reroute.
...
Thanks, this was very informative
Btw, you can click the star icon to follow a thread without commenting
hls-b1 has been disconnected for our AS once again.
Everything was swapped on that link, it worked for a few days before issues began again. No more packet loss at all, but throughput is very low once again.