[GH-ISSUE #2645] Recursor: Retransmission policy #1038

New issue

Closed

opened 2026-03-16 01:23:29 +03:00 by kerem · 1 comment

kerem commented

2026-03-16 01:23:29 +03:00

Owner

Originally created by @divergentdave on GitHub (Dec 5, 2024).
Original GitHub issue: https://github.com/hickory-dns/hickory-dns/issues/2645

The recursor should have some retransmission policy, in order to gracefully handle packet loss when using UDP. Section 4.2.1 of RFC 1035 has some suggestions.

Note that, while parallel_conn_loop() does have some logic to retry with an exponential backoff, it is only triggered by ProtoErrorKind::Busy. This error is only generated when internal channels are full in DnsMultiplexer or BufDnsRequestStreamHandle.

Originally created by @divergentdave on GitHub (Dec 5, 2024). Original GitHub issue: https://github.com/hickory-dns/hickory-dns/issues/2645 The recursor should have some retransmission policy, in order to gracefully handle packet loss when using UDP. Section 4.2.1 of RFC 1035 has some suggestions. Note that, while `parallel_conn_loop()` does have some logic to retry with an exponential backoff, it is only triggered by `ProtoErrorKind::Busy`. This error is only generated when internal channels are full in `DnsMultiplexer` or `BufDnsRequestStreamHandle`.