Pv6 connectivity issues from NRT region to Cloudflare Workers (ongoing since 2025-10-17)

[nappa0326]

We are experiencing persistent IPv6 connectivity issues from our Fly.io app (deployed in NRT region) to Cloudflare Workers since around October 17, 2025, which correlates with the https://status.fly.io/.

Environment:

• Region: nrt (Tokyo/Narita)
• Platform: Node.js 18-alpine
• Target: Cloudflare Worker (*.workers.dev domain)

Symptoms:

1. All outbound HTTPS connections to Cloudflare Workers time out after 10-30 seconds
2. netstat shows connections stuck in SYN_SENT state with IPv6 addresses
3. Direct S3-compatible API access (bypassing Worker) works fine via IPv4

Evidence:

SSH into machine

fly ssh console

Test connection to Worker

curl -v https://[worker-domain].workers.dev/health

Result: Operation timed out after 30000 milliseconds

Check connection state

netstat -an | grep 2606:4700

Shows: SYN_SENT state (connection never established)

Status Page Note:
The official https://status.fly.io/ mentions only the SJC region, but our NRT-region app is clearly affected.

Workaround Attempted:

• NODE_OPTIONS=“–dns-result-order=ipv4first” → No effect
• undici Agent with IPv4 forcing → No effect
• Bypassing Worker entirely → Works (confirms IPv6 issue)

Questions:

1. Is NRT region also affected by the IPv6 incident?
2. Is there an ETA for a fix?
3. Should we consider disabling IPv6 entirely for our app as a permanent solution

[lillian]

hi, this is an unrelated issue for one single worker host in nrt region. we’ve put up a host-specific issue (you should see the message in your dashboard now) and disabled new machine provisions on the host, so deleting and recreating your machines should get you back up and running.

I would not recommend disabling IPv6 permanently, no, that will break more things than it fixes.