search

cilium / cilium-olm

Other
10 stars 13 forks source link

OpenShift 4.9 #49

Closed errordeveloper closed 3 years ago

errordeveloper commented 3 years ago

It appears that Cilium is not working on 4.9 development builds (see https://github.com/openshift/release/pull/18681#issuecomment-903672655).

There are 4.9.0-fc.0 binaries already available, and it should be easy to test this also with https://github.com/cilium/openshift-terraform-upi/commit/c87d64c7b011480cd3940a1b00743f151b38d111.

errordeveloper commented 3 years ago

This is quite likely to do with deprecation of apiextensions.k8s.io/v1beta1 in Kubernetes as OpenShift 4.9 is based on 1.22.

Interestingly enough this project originally started with v1, and v1beta1 had to be used due to issue with the certification system (see #23). It turns out that v1 was already supported in OpenShift 4.4.

errordeveloper commented 3 years ago

xref https://github.com/openshift/release/pull/21293

errordeveloper commented 3 years ago

This is partly blocked by https://bugzilla.redhat.com/show_bug.cgi?id=1997050, but without_kube_proxy: true gets Cilium running.

errordeveloper commented 3 years ago

CRD version was addressed in https://github.com/cilium/cilium-olm/commit/47ff4dfd32ef7c17fe95ed6c4feff8a6bb937218, and the certification system should accept v1 now.

errordeveloper commented 3 years ago

There was a minor issue with ports, seemingly port 8080 was taken by another process https://github.com/cilium/cilium-olm/commit/d8c642682fb33a83599a2b3e477924b4cb1d7759.

errordeveloper commented 3 years ago

https://github.com/openshift/cluster-network-operator/pull/1188

errordeveloper commented 3 years ago

openshift/cluster-network-operator#1188

This has just been merged.

vrutkovs commented 3 years ago

https://github.com/openshift/release/pull/21293 merged

errordeveloper commented 3 years ago

I think this can be closed pending it's validated against a 4.9 build that include CNO fixes.

@vrutkovs https://github.com/openshift/cluster-network-operator/pull/1188 was merged on 01/09/21, and 4.9.0-fc.1 appears to be released on that day also. How can I check which version of CNO was shipped in 4.9.0-fc.0 (or a latest nightly build) without deploying a cluster?

vrutkovs commented 3 years ago

How can I check which version of CNO was shipped in 4.9.0-fc.0 (or a latest nightly build) without deploying a cluster?

$ oc adm release info quay.io/openshift-release-dev/ocp-release:4.9.0-rc.0-x86_64 --commit-urls | grep cluster-network-operator
  cluster-network-operator                       https://github.com/openshift/cluster-network-operator/commit/8437b077d5700f3b4f484f34717c939faf90c5e2

Yup, its in rc.0 (and latest nightlies too)

errordeveloper commented 3 years ago

@vrutkovs ah, I didn't realise rc.0 was out already, thanks very much!

errordeveloper commented 3 years ago

I've started deploying 4.9.0-rc.0 just now, and can see that all issues discussed here had been fixed - control plane nodes and Cilium are all running happily.

$ kubectl --kubeconfig tfc-scripts/test-490rc0-oss1103-1.kubeconfig get nodes
NAME                                        STATUS   ROLES    AGE   VERSION
ip-10-0-50-199.eu-west-1.compute.internal   Ready    master   60s   v1.22.0-rc.0+75ee307
ip-10-0-61-143.eu-west-1.compute.internal   Ready    master   88s   v1.22.0-rc.0+75ee307
ip-10-0-61-169.eu-west-1.compute.internal   Ready    master   88s   v1.22.0-rc.0+75ee307
$ kubectl --kubeconfig tfc-scripts/test-490rc0-oss1103-1.kubeconfig get pods -n cilium
NAME                               READY   STATUS    RESTARTS   AGE
cilium-88xxl                       1/1     Running   0          56s
cilium-j7k2v                       1/1     Running   0          56s
cilium-olm-66b748845-hslfc         1/1     Running   0          4m22s
cilium-operator-588fb5ff4b-2lt5t   1/1     Running   0          56s
cilium-operator-588fb5ff4b-46qm5   1/1     Running   0          56s
cilium-xpg5f                       1/1     Running   0          56s
$ kubectl --kubeconfig tfc-scripts/test-490rc0-oss1103-1.kubeconfig get pods -n openshift-network-operator
NAME                               READY   STATUS    RESTARTS   AGE
network-operator-f889465bb-zs7wl   1/1     Running   0          5m53s
$

If something else happens and bootstrap doesn't complete in my test, I'll open another issue to dig into that.

errordeveloper commented 3 years ago

I definetly worked:

$ kubectl --kubeconfig tfc-scripts/test-490rc0-oss1103-1.kubeconfig get clusterversion
NAME      VERSION      AVAILABLE   PROGRESSING   SINCE   STATUS
version   4.9.0-rc.0   True        False         86m     Cluster version is 4.9.0-rc.0
$ kubectl --kubeconfig tfc-scripts/test-490rc0-oss1103-1.kubeconfig get pods -n cilium
NAME                               READY   STATUS    RESTARTS       AGE
cilium-675hm                       1/1     Running   0              106m
cilium-88xxl                       1/1     Running   0              114m
cilium-bk6ls                       1/1     Running   0              103m
cilium-j7k2v                       1/1     Running   0              114m
cilium-lbl6h                       1/1     Running   0              106m
cilium-olm-7dff7dd48d-tz47h        1/1     Running   2 (105m ago)   112m
cilium-operator-588fb5ff4b-2lt5t   1/1     Running   1 (111m ago)   114m
cilium-operator-588fb5ff4b-46qm5   1/1     Running   1 (105m ago)   114m
cilium-xpg5f                       1/1     Running   0              114m
$

(cilium-operator and cilium-olm restarts are known issues during early bootstrap stage).