Resolution providing incomplete answer on CoreDNS 1.6.6 #4563
-
Hello all! I'm trying to diagnose why resolving a domain like "foo.cognitiveservices.azure.com" is giving me a different result in our K8s cluster than outside. Our cluster is an EKS cluster on Kubernetes 1.17 using coredns Since I cannot reproduce this on a later version of CoreDNS, I assume there is either a different setting or a bug fixed between the two versions. For example:
And without coredns:
I ran a tcpdump to capture the request and the response. The requests are functionally identical (both querying an A record), however the response from CoreDNS appears to be one record short of the answer. Indeed, if I query tl;dr: Is there a setting that we can change to resolve this, or are we looking at upgrading our cluster to a newer version to resolve this? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 3 replies
-
There was a truncation bug in 1.6.6 where the forward plugin would unset the truncation bit of a truncated response from an upstream before sending the response to the client. It was fixed upstream, but it didn't make it into CoreDNS until 1.8.0. So that doesn't explain why it seems to work in 1.7.1. |
Beta Was this translation helpful? Give feedback.
There was a truncation bug in 1.6.6 where the forward plugin would unset the truncation bit of a truncated response from an upstream before sending the response to the client. It was fixed upstream, but it didn't make it into CoreDNS until 1.8.0. So that doesn't explain why it seems to work in 1.7.1.