cloudflare / cloudflare/cloudflared
UDP payload exceeds transport MTU 1280
- Dominant language
- Go
- Stars
- 15.6k
- Forks
- 1.4k
- PR merge metrics
- No merged PRs in 30d
Description
Hello,
When running cloudflared on bare metal (or docker), attempts to pass certain UDP traffic via the WARP client results in the log being spammed (multiple times per second) with the following:
`ERR Failed to send session payload from destination to transport error="origin UDP payload has 1440 bytes, which exceeds transport MTU 1280"`
Further to this, performance of the application is badly affected.
Modifying the MTU (to 1280, 1340, etc) directly on the network interface serving cloudflared (and ensuring ICMP messages are not blocked between the origin and cloudflared) results in a change of error message:
`ERR Failed to send session payload from destination to transport error="Failed to send datagram back to edge: DATAGRAM frame too large"`
This issue is mitigated when using http2 instead of quic, however that also breaks DNS (via WARP: Local DNS Fallback / Resolver Policies) and ICMP, as quic is a requirement for those to work.
Can you offer any advice, please?
Regards
Contributor guide
Assessment
This issue has not been assessed yet.