There is right now only 5 IP4 ntp servers left in the pool now and it is only a matter of time before my own server is also pushed out because it recieves above 100Mbit/s
It is down to one host now.
Looks like it is back up to 11/20…
We provide two to the CN pool, and they’ve had scores around 20 throughout the last 24 hours. I couldn’t believe the number in the CN pool is down this low right now (5 at time of posting).
I can only conclude that monitoring is absolutely screwed, and no progress has been made on matters discussed in Adding servers to the China zone and other threads (e.g. monitoring CN servers that are hosted behind the GFW with a monitoring station behind the GFW). I get that this is a volunteer project, but while the problems continue, we’ve got around a dozen volunteers propping up the zone by pulling in the order of 1Mpps of traffic around the globe.
Down to 4 now and I am receiving 100+Mbps. The pattern is repeating again today and my host is lucky if it “survives” the next check by the monitor, since about 20% of the network traffic is thrown away now.
As I said before, allowing A+AAAA hybrid reply from *.pool.ntp.org (not only 2.pool) shall be able to help ease the burden more or less. Don’t know why this cannot be implemented… situation for most Africa/Asia/south America zones cannot be worse off.
Hm - I could be convinced to implement this first for China since it’s struggling getting it to work well anyway… I thought that most users there still don’t have IPv6?
(I’ve been making progress on fixing up how the monitoring metrics/logs are handled so more monitoring systems can be in production).
Any idea of timescales?
Chinese government issued some executive orders demanding ISPs to provide IPv6 to their end users during 2017-2018. As of the start of 2019, users of major ISPs already have their native IPv6 addresses allocated automatically.
There are many that still don’t get IPv6 access, and the provided IPv6 links in some regions have bad connectivity, especially to international networks. This partially due to limited backbone capacity and / or exchange bandwidth between ISPs, but it’s already way more better than nothing, and we can except it to improve in near future.
So I’d say, providing AAAA response to clients in CN pool is already practicable. There might be new questions like whether ISPs implemented QoS on their IPv6 backbones for UDP packets, but we won’t know them if we don’t really run traffic on IPv6.