1511865 - Intermittent DNS resolution failures

Status: RESOLVED INVALID

(Firefox :: Untriaged, defect)

Description

[Yuval Adam]

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:63.0) Gecko/20100101 Firefox/63.0

Steps to reproduce:

On occasion, DNS resolution to certain domains will fail via Firefox, while system DNS resolution works to the same domains works just fine.

1. Navigate to a website (happens a lot on Google and AWS domains for some weird reason) and receive "Server not found"
2. Verify DNS lookup via `about:networking#dnslookuptool` yields NS_ERROR_UNKNOWN_HOST
3. Verify system resolution to the exact same domain works via `dig` and `host` command line tools.
4. Atempt steps 1 + 2 again with same results
5. Only after a few minutes problem resolves organically (until the next time)



Actual results:

DNS lookup fails with NS_ERROR_UNKNOWN_HOST while system lookup on same domain succeeds.


Expected results:

Successful domain lookup

Comment 1

[Yuval Adam]

Attachment: log.txt-main.1247

Comment 2

[Yuval Adam]

I added a log file showing the nsHostResolver behavior on one sample AWS domain.

Also for debugging purposes I set network.dns.disablePrefetch and network.dns.disableIPv6 both to true, and problem still persists.

Comment 3

[Yuval Adam]

Looks like systemd-resolved is spitting lots of errors to the journal about DNSSEC:

Sep 21 23:00:39 host systemd-resolved[582]: DNSSEC validation failed for question eu-west-1.console.aws.amazon.com IN AAAA: no-signatu>
Sep 21 23:00:39 host systemd-resolved[582]: DNSSEC validation failed for question eu-west-1.console.aws.amazon.com IN A: no-signature
Sep 21 23:00:39 host systemd-resolved[582]: DNSSEC validation failed for question eu-west-1.console.aws.amazon.com IN SOA: no-signature
Sep 21 23:00:39 host systemd-resolved[582]: DNSSEC validation failed for question eu-west-1.console.aws.amazon.com IN DS: no-signature

This doesn't explain other domains that fail on occasion (such as Google) but until I have further evidence I'm going to close this bug. Sorry for the false alarm.

Comment 4

[Yuval Adam]

For posterity, this is likely the core issue - https://github.com/systemd/systemd/issues/9771