starred/hickory-dns

Fork 0

mirror of https://github.com/hickory-dns/hickory-dns.git synced 2026-04-25 19:25:56 +03:00

[GH-ISSUE #612] Query multiple nameservers in parallel #252

New issue

Closed

opened 2026-03-07 23:01:51 +03:00 by kerem · 5 comments

kerem commented

2026-03-07 23:01:51 +03:00

Owner

Originally created by @stuartnelson3 on GitHub (Nov 11, 2018).
Original GitHub issue: https://github.com/hickory-dns/hickory-dns/issues/612

Is your feature request related to a problem? Please describe.
Querying multiple nameservers in parallel is something done by musl to improve performance, cf. https://wiki.musl-libc.org/functional-differences-from-glibc.html#Name-Resolver/DNS.

My feature request is to allow this functionality.

Describe the solution you'd like
Allow setting parallel lookups in ResolverOpts, and limiting the number of parallel lookups at a time. For example, maybe I only want have 3 parallel requests at once (a la musl). Related to #606, I can further say to take the first response that isn't a failure, or if all fail, returning one of those. This requires some thought as a response is then only as fast as it's slowest failure (or timeout).

Describe alternatives you've considered
If a SERVFAIL didn't abort trying other nameservers, that would be an alternative, albeit trying requests in serial is not ideal typically.

Additional context
See #606

Originally created by @stuartnelson3 on GitHub (Nov 11, 2018). Original GitHub issue: https://github.com/hickory-dns/hickory-dns/issues/612 **Is your feature request related to a problem? Please describe.** Querying multiple nameservers in parallel is something done by musl to improve performance, cf. https://wiki.musl-libc.org/functional-differences-from-glibc.html#Name-Resolver/DNS. My feature request is to allow this functionality. **Describe the solution you'd like** Allow setting parallel lookups in ResolverOpts, and limiting the number of parallel lookups at a time. For example, maybe I only want have 3 parallel requests at once (a la musl). Related to #606, I can further say to take the first response that isn't a failure, or if all fail, returning one of those. This requires some thought as a response is then only as fast as it's slowest failure (or timeout). **Describe alternatives you've considered** If a SERVFAIL didn't abort trying other nameservers, that would be an alternative, albeit trying requests in serial is not ideal typically. **Additional context** See #606

kerem

2026-03-07 23:01:51 +03:00

closed this issue
added the
enhance

perf

crate:resolver
labels

kerem commented

2026-03-07 23:02:07 +03:00

Author

Owner

@bluejekyll commented on GitHub (Nov 11, 2018):

FYI, I’ve got a patch coming for #606. This one, we’ll have to consider options on. I think we might want to have a maximum on the number of parallel queries, and default to 2 or so.

@bluejekyll commented on GitHub (Nov 11, 2018): FYI, I’ve got a patch coming for #606. This one, we’ll have to consider options on. I think we might want to have a maximum on the number of parallel queries, and default to 2 or so.

kerem commented

2026-03-07 23:02:12 +03:00

Author

Owner

@stuartnelson3 commented on GitHub (Nov 12, 2018):

Makes sense. It might be easier to say max number of nameservers. If we allow more nameservers than parallel attempts, what is the behavior if all nameservers in the N that get queried fail? Continue trying with the next batch, or fail, and start priority sorting nameservers...

I think musl's policy makes sense: set a max number of nameservers to be queried.

@stuartnelson3 commented on GitHub (Nov 12, 2018): Makes sense. It might be easier to say max number of nameservers. If we allow more nameservers than parallel attempts, what is the behavior if all nameservers in the N that get queried fail? Continue trying with the next batch, or fail, and start priority sorting nameservers... I think musl's policy makes sense: set a max number of nameservers to be queried.

kerem commented

2026-03-07 23:02:17 +03:00

Author

Owner

@bluejekyll commented on GitHub (Nov 12, 2018):

Perhaps both a max nameserver attempts and max parallel requests would be best? There already exists a max_retries option so we should consider its relationship to these new options.

I haven’t really considered exactly how to implement this yet, but we should be able to pop N nameservers from the set of M in the pool, wher N is the number to execute in parallel. Then in the case where no nameserver returned a successful result, use the existing loop to continue popping N until we’ve attempted all M, unless we hit the max_attempts as described here.

@bluejekyll commented on GitHub (Nov 12, 2018): Perhaps both a max nameserver attempts and max parallel requests would be best? There already exists a max_retries option so we should consider its relationship to these new options. I haven’t really considered exactly how to implement this yet, but we should be able to pop N nameservers from the set of M in the pool, wher N is the number to execute in parallel. Then in the case where no nameserver returned a successful result, use the existing loop to continue popping N until we’ve attempted all M, unless we hit the max_attempts as described here.

kerem commented

2026-03-07 23:02:22 +03:00

Author

Owner

@stuartnelson3 commented on GitHub (Nov 12, 2018):

Sounds reasonable to me

@stuartnelson3 commented on GitHub (Nov 12, 2018): Sounds reasonable to me

kerem commented

2026-03-07 23:02:28 +03:00

Author

Owner

@bluejekyll commented on GitHub (Nov 14, 2018):

PR #615 is the current proposed fix for this, I think it's simple enough to incorporate. Right now it defaults to 2 concurrent requests.

@bluejekyll commented on GitHub (Nov 14, 2018): PR #615 is the current proposed fix for this, I think it's simple enough to incorporate. Right now it defaults to 2 concurrent requests.

kerem referenced this issue

2026-03-15 22:15:17 +03:00

[GH-ISSUE #190] resolver lookup_ip should checkout /etc/hosts/ before dns requests? #387

kerem referenced this issue

2026-03-15 22:24:01 +03:00

[GH-ISSUE #252] Support Windows /etc/hosts file when system configuration is used? #416

kerem referenced this issue

2026-03-16 02:13:16 +03:00

[PR #623] [MERGED] Support Windows hosts file when system configuration is used #1562