This message was deleted.
# general
a
This message was deleted.
c
Update: connecting to a node and checking /etc/rancher/rke2/registries.yaml
{"configs":{"harbor-registry":{"auth":null,"tls":{"ca_file":"","cert_file":"","key_file":"","insecure_skip_verify":true}}},"mirrors":null}
-> insecure_skip_verify is true the regarding image is available, no auth required. pulling from outside the cluster works. curl from a rke2 node with "--insecure" works.
a
We ran into something similar a few weeks ago when a department that uses a downstream Rancher cluster deployed a private registry within it. The solution was to add the cert to the trust store on the underlying hosts (Ubuntu 22.04 on VMWare ESX, in our case). This was just for short-term use, as the department is moving to a cloud-hosted registry, but needed an interim solution for testing. An example of a scalable approach for adding the certs to the trust store is outlined in step 1 of this blog post (of note, we did not do this, but would look at a similar approach for long-term needs): https://www.linuxtechi.com/setup-private-docker-registry-kubernetes/ And, the underlying reason for having to do this appears to be a current limitation of containerd, which may be resolved in a 2.x release. More details here: https://stackoverflow.com/questions/53545732/how-do-i-access-a-private-docker-registry-with-a-self-signed-certificate-using-k/72216347#72216347
c
Is your image literally at
harbor-registry/repo/image:tag
?
You need to have a dot or semicolon in the registry name, otherwise
harbor-registry
will be parsed as part of the repo name on docker hub, not the registry hostname.
That is just how image references work
c
Hello! Thanks for your replies! docker pull harbor-registry.domain.tld/project/repository:latest -> works deploying the app from this repository via ArgoCD to Openshift works. Our harbor
Hello! Thanks for your replies! think I got the issue: tried to manually add the cacert:
root@cluster07-pool1-d735bac3-shj2x:~# update-ca-certificates
Updating certificates in /etc/ssl/certs...
rehash: warning: skipping ca-certificates.crt,it does not contain exactly one certificate or CRL
rehash: warning: skipping private-ca.pem,it does not contain exactly one certificate or CRL
1 added, 0 removed; done.
Running hooks in /etc/ca-certificates/update.d...
done.
Looks like it doesn't like our ca certificate chain (1 root, 2 sub-ca's). Will have a look into this.
I've verified that the ca certs are correct - they are included in the /etc/ssl/certs/ca-certificates.crt file on the rke2 nodes. openssl s_client -CAfile ca.pem -connect rancher.yourdomain.com:443 also works. Background: we're using our internal enterprise pki and roll out the ca certificates on all servers - including the docker-vm where rancher itself is running. Same CA is used by our internal harbor registry. Since there are no issues on other systems when pulling from there, there has to be a configuration problem on the rancher-deployed rke clusters. I've got no idea why they can't pull...
It works when deploying the CA certs via cloudinit. but it feels more like a workaround...
👍 1