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.
avoro network packet loss
https://imgur.com/a/lPIpz5z today or a month ago it's the same
depending on the speed you can get loss almost every second, it's mostly inbound, and no, it's not because i overload the network by myself, it can handle 1gbit/s but i never exceed 100mbit/s
a good example for reference: https://imgur.com/a/nFdv81F
Comments
Did you open a ticket about it?
@lukast__ yes, reported it for the first time 30 days ago, they basically said it looks fine from their side, 7 days ago i asked them to report it to their network engineer and im still waiting for a response.
also it looks like today they blocked large ICMP packets >954(982) to some networks including my oracle vps which i used for testing
https://imgur.com/a/MUd4Rvb
Try open the ticket and have them turn off the anti-ddos for your server ip
good idea. solves some problems with my vps too.
@dataforest: You said:
But that doesn't seem possible or where can I do that?
And @0ka, ICMP isn't suited well for that, @dataforest said that they shape it, so it can be slow/get lost.
they disabled it and large icmp now works fine, thanks. btw i didn't enable it by myself and didn't ask anything about it
@lukast__ icmp/tcp/udp they all have loss, even after they disabled the protection, download speed fluctuates a lot right now in the evening (4-10mb/s)
https://imgur.com/a/RRAS4OE
eh... disable anti-ddos normaly takes 24h...
But that could also simply be peering.
my smokeping with 1400 byte packets started right after i got a response from support
https://imgur.com/a/qQwYu94
mtr from oracle to avoro reports losses only on the last hop, mtr from avoro to anything reports losses on the first hop already (https://imgur.com/a/7SzAvSa). also my first screenshot of mtr was taken 10 days ago, the route looked like this then: oracle-twelve99-cdn77-synlinq-avoro, today it's like this: oracle-twelve99-gcore-zerolag-avoro (https://imgur.com/a/eN2gFLB) and loss is still there, it's something in their internal network, drivers or virtualization software
udp: updated screenshots for better understanding
Hi,
please note, that we cant offer support via Lowendtalk. I tested Oracle from our ansible: https://imgur.com/T1XEGVT - also (in your support ticket) there was 0,1 % loss on about 27000 packets, I dont know what it was, but it can also come from the filter which was disabled within 17 minutes today.
Your VPS is currently on its own node, there is no other server on it, there are no malfunctions or general loss on our part. If there is a fundamental problem with routing (e.g. to Oracle), our support team will be happy to help and we will check whether the traffic can be rerouted, but with 0.1% loss at 22,000 packets, with the filter activated, I don't see any need for this. Especially as we are already seeing (see our screenshot) a 0.5% loss at Oracle.
@dataforest thanks for responding, but it's not only oracle, even 1.1.1.1 is lossy, can you please test on your machine with this "wget -O /dev/null https://speedtest.selectel.ru/100MB"? the speed is very unstable ~8mb/s, it can reach 80mb/s on another vps providers
@dataforest also i updated my post which is above yours, not sure you saw the updated version
Of course:
root@ansible: ~ # wget -O /dev/null https://speedtest.selectel.ru/100MB
--2024-06-30 20:12:49-- https://speedtest.selectel.ru/100MB
Auflösen des Hostnamen »speedtest.selectel.ru (speedtest.selectel.ru)«... 2a00:ab00:600:1::5, 188.93.16.211
Verbindungsaufbau zu speedtest.selectel.ru (speedtest.selectel.ru)|2a00:ab00:600:1::5|:443... verbunden.
HTTP-Anforderung gesendet, warte auf Antwort... 200 OK
Länge: 104857600 (100M) [application/octet-stream]
In »»/dev/null«« speichern.
/dev/null 100%[==============================================================================================================================================>] 100,00M 38,6MB/s in 2,6s
I also installed a test VPS on the same node you are and run the MTR: https://imgur.com/a/6ppr8DB
@dataforest you got a better result than i get right now https://imgur.com/a/H1ucgKI
@dataforest i used this command for testing: "mtr -s 1472 -i 0.3 1.1.1.1"
We have a big peering with Cloudflare in FRA, It won't be more direct, the port is currently at 20% utilization, so everything is fine. The interconnect between the data center and Interxion (where the traffic arrives in this case) is also at 25%.
@dataforest https://imgur.com/a/lOgYUVZ
BTW BIG thumbs up for looking into this
upd: my vm kernel panicked and locked up, reboot/stop/shutdown on the website return an error, this happened to me before and only tech support could reboot my vm
@dataforest where is that interface you mentioned?
@dataforest the closer to the night, the better the result.
try this tomorrow on the node and in the vm at the same time, good luck and thanks.
"sleep $((60 - $(date +%s) % 60)); while true; do echo -n $(date +%R); ping -W 1 -s 1472 -i 0.1 -c 600 1.1.1.1 | grep loss | awk -F',' '{print $3}'; done"
https://imgur.com/a/dgAjquY
it will wait for a new minute to start before doing measurements.
Hi,
seems fine:
root@ansible: ~ # sleep $((60 - $(date +%s) % 60)); while true; do echo -n $(date +%R); ping -W 1 -s 1472 -i 0.1 -c 600 1.1.1.1 | grep loss | awk -F',' '{print $3}'; done
22:38 0% packet loss
22:39 0% packet loss
22:40 0% packet loss
22:41 0% packet loss
22:42 0% packet loss
22:43 0% packet loss
Traffic is at its highest level on our Cloudflare PNI - if you're not happy we can offer you a refund, for that please answer the ticket in our Control Panel
@dataforest is this a different person? im not sure if you're reading my messages completely. The loss is present on any ip address, 1.1.1.1 and my screenshot above is just an example. Im almost happy and don't want a refund, you provide great services, but i want to help and resolve this issue.
Can you set up an iperf3 server somewhere in your network? it may make it clearer to understand where the issue might be, with your report a few msgs above i can see that on the node itself the network works better.
And can i have a direct contact with someone with access to the node?
thanks
Hi,
do you have Telegram? If yes you can contact me there, so we can check it together.
i got a second test vm on the same node and this is really strange
https://i.imgur.com/ZHsRgO7.png
vm on the right has constant packet loss, anyone has a clue what's wrong? i did iperf from multiple sources and the result is the same