Server ID: server-69e673fe214f35a455c92809
Plan: Aliyun Hong Kong 2C/4GB/200M $8/month
Instance type: ecs.e-c1m2.large
Issue:
The plan is labeled as "200M" (interpreted as 200 Mbps public bandwidth), but the actual download speed is only 0.13–0.92 Mbps (285 times slower than the advertised value).
Evidence:
- speedtest-cli (multiple local HK servers)
- 1010 (HK): Ping 2.9 ms, Download 0.14 Mbps
- SmarTone (HK): Ping 179 ms, Download 0.19 Mbps
- Auto-select: 0.70-0.13 Mbps, fluctuating across multiple tests
- Pulling Zeabur registry (registry-oci.zeabur.cloud) in the same region
- Ping: 1.97 ms
- 60 seconds to download image layer only completed 6.6 MB (~0.92 Mbps)
- ECS machine capability check
- /sys/class/net/eth0/speed: -1 (no info for virtual NIC)
- max-netbw-egress (Aliyun metadata): 200 Mbps
- Pulling mirror.cloud.aliyuncs.com via VPC intranet: 1.6 MB/s = 12 Mbps (limited by mirror side, not the machine)
- Control group: Mac pulling the same image layer from home network
- 81.7 Mbps, finished 202 MB in 20 seconds
- Proves the registry is not the bottleneck
- No rate limits set in local iptables / tc qdisc
Suspicion: EIP public downlink bandwidth is throttled to 1 Mbps (despite the 200M plan label).
Request:
- Please confirm the actual EIP public bandwidth configuration for server-69e673fe214f35a455c92809.
- If there is a configuration error, please fix it.
- If "200M" refers to something else (e.g., Aliyun ECS NIC baseline), please clarify.
Impact: Because the server OOMed today, the K3s image cache was GC'd. All images need to be re-pulled from the registry. A 730 MB image takes 2+ hours at 0.7 Mbps.