Closed der-eismann closed 1 year ago
Similarly, if you deploy coredns
via a helm chart, you end up with a deployment named something like coredns-coredns
(or RELEASE-NAME-coredns
). This also bypasses the check here, and so makes traefik-mesh not functional.
https://github.com/traefik/mesh/blob/e53767daee443f55338a4ffcf07e66b39afb630f/pkg/dns/client.go#L85
Further, it makes mitigating #787 require deploying coredns directly (i.e. not using helm)
Maybe at a minimum making this configurable via an env var / etc. (and then exposing as a value in your helm chart) would give us a workaround to point at deployments / deamonsets, other names, etc.
Is there anyway to bypass this check ? I am on a managed kubernetes cluster and the image name tag used for coredns deployment is a custom tag without containing exact version string that code is looking for
parts := strings.Split(container.Image, ":")
Would it be possible to provide a way to consciously avoid this check is such cases ?
Closing because it's been 2 years now and we don't use this anyway
Feature Request
Proposal
I'd like traefik mesh to also support CoreDNS when used as a daemonset instead of a deployment.
Background
We use one dns pod per worker to achieve a lower latency, however in the code it explicitly looks for a deployment which must be named
coredns
. A bit more flexibility on that part would be great.Workarounds
None that I know of.