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.
BuyVM - Slow speeds on the servers in Bissen/Luxembourg
Anthony2016
Member
in Providers
Is the speed on your servers in Bissen (Luxembourg) just as bad:
https://www.speedtest.net/result/14707190695
https://www.speedtest.net/result/14707226815
Please don't be surprised by the different names ("Freemesh Québec" & "INTERCLOUD UK"): We use our own subnet we brought with us.
I would be in favor of introducing monthly traffic limits of 30 TB. That would certainly improve the situation. What do you think of this proposal?
Monthly traffic limits of 30TB on all BuyVM machines
- Are monthly traffic limits of 30TB on all BuyVM machines a73 votes
- ...good idea50.68%
- ...bad idea49.32%
Comments
It's been almost 2 hours and I'm still downloading the same 3GB file, so yeah I feel you lol
Limit then reduce speed(100mbps) is ok...
Limit then block is not ok.
BuyVM isn't 100% unmetered. FUP is 25 Mbps per 1GB of RAM for the 95th percentile.
Speeds from a 1g box, you probably need to specify the correct location/select the speedtest server closer to you.
Yes, as bad as yours. I'm even going to give up on renewing it.
Do you have a ticket open about it?
We have a switch that's maxing it's port pretty bad (the 5900x's) so I think that's likely where you are. LUX as a whole is only using ~40gbit of the 100gbit HE port it has.
We're waiting on another 100G switch to allow us to bond some interfaces, but i'll probably just spread some users out.
We have been capping very heavy users as needed as well.
Francisco
For what it's worth it's either with cegecom, the transport provider, or hurricane. I was thinking it was one of our switch ports having a bad optic or just 40G not working like I was expecting (was seeing it max around 25 - 30G), but internal iperf's show that it isn't that.
A look at our weekly graphs shows a sudden drop in speed, and bouncing a little traffic to Cogent, proved it too.
I've logged tickets with both of them so we'll see who bites first.
Francisco
For those that are complaining about bandwidth.
Keep in mind fair-use.
So as far as @Anthony2016 proposal. There is sometimes more to the story than..."I got slow speeds."
Open a ticket with BuyVM and then include MTR or Traceroute as it can show where the bottleneck is.
Sometimes it is the provider your trying to connect to.
nah nah nah. This is a network fault, our graphs show it.
I'm doing some changes at the moment to see if I can fix our Cogent port which will free up a lot of capacity on the HE side, allowing them/cege to sort things.
Francisco
Looks better now!
It's not really running smoothly yet.
Since switching all outgoing traffic in Bissen from HE to Cogent, we've been experiencing short-term outages all the time. It's not possible to listen to a 128 kbps web radio stream without dropouts.
I really hope you have a ticket open for that and put all that info there too for @Francisco to work this shit out and you are just posting here in parallel to keep community informed.
Of course I opened a ticket. I wonder if others are having the same problems with the Luxembourg server.
I got a bunch of Lux boxes, I'll check when back at my desk. I've not noticed any issues, though admittedly most aren't monitored for performance as they're dev or project stuff. The production ones that I do monitor and care about haven't had any issues that I'm aware of, however.
It's all those tor relays taking up a lot of bandwidth
According to my information, the uplink from HE continued to cause problems, which is why all traffic (incoming and outgoing) is now routed via Cogent.
If Cogent stays stable then everything is fine.
No. We're only around 25% utilization normally in LUX (around 40 - 50Gbit of 200Gbit online).
Our 100G port to HE was on the fritz earlier in the week. We reached out to HE who reported a metric crap load of CRC errors. We replaced the optic on our end just to be safe, no improvement. This morning the port starts to constantly drop for a few seconds, but enough to kill BGP, that's why peoples sessions dropped.
For now we've offloaded things to Cogent and hope to have HE humming along again ASAP. I've been spamming Cogecom's email hoping to get a reply soon. Worst case I'll call them in the morning.
Given what we've seen on the line in the past little bit I'm thinking something really blew up on their end.
EDIT - I should add, we originally thought it was an issue on a switch, but that didn't add up given we still had 15Gbit or so spare on each port. It's only once I started shifting routes towards some of the speedtest servers did I see the issue was only happening on the HE link, and only with uploading (buyvm -> them).
Francisco
Same here, my server is barely usable now
>
The HE port is turned down completely at the moment, but maybe that's where your issue is. All depends on how your ISP is working with Cogent. DTAG is a dumpster fire I think, not sure on others.
Francisco
In this moment, it works better than it was hour ago. And yeah.. I'm on DTAG network (T-Mobile Poland)
Cogent seems to be working well.
Working fine for me
It looks like HE is available again, so I'm assuming the issue is resolved.
@Francisco, Any eta on Luxembourg storage slabs re-stock?
Whenever there's cancellations I guess
Francisco> @Anthony2016 said:
Yes. Cogecom had a bad optic on our link to HE. The link is 'active' and not passive so it made it hard to debug. They ended up doing a full line test to address it. Thankfully it's resolved now.
Francisco