Current Example needs at least 3 worker nodes

[kube-vip]

Current Example needs at least 3 worker nodes

[thebsdbox]

Ah ha! It certainly looks like that. The strategy behind this is that you need a single node... anything else provides HA and the rules in the deployment mean that we can’t have overlapping pods. It won’t fail.. it will just be left pending which isn’t an issue :-)

[voor]

While that is certainly the case for kubectl, more applications (in particular I used kapp) try to be a little more helpful during the installation process, so it'll hang and output the logs that indicate not all pods can be appropriately scheduled (this is a 3 master node, 1 worker node example):

8:48:12PM: ongoing: reconcile deployment/kube-vip-cluster (apps/v1) namespace: default
8:48:12PM:  ^ Waiting for 2 unavailable replicas
8:48:12PM:  L ok: waiting on replicaset/kube-vip-cluster-76c6cfdd77 (apps/v1) namespace: default
8:48:12PM:  L ongoing: waiting on pod/kube-vip-cluster-76c6cfdd77-mrxnb (v1) namespace: default
8:48:12PM:     ^ Pending: Unschedulable (message: 0/4 nodes are available: 1 node(s) didn't match pod affinity/anti-affinity, 3 node(s) had taints that the pod didn't tolerate.)
8:48:12PM:  L ongoing: waiting on pod/kube-vip-cluster-76c6cfdd77-8cvvr (v1) namespace: default
8:48:12PM:     ^ Pending: Unschedulable (message: 0/4 nodes are available: 1 node(s) didn't match pod affinity/anti-affinity, 3 node(s) had taints that the pod didn't tolerate.)
8:48:12PM:  L ok: waiting on pod/kube-vip-cluster-76c6cfdd77-5mkpb (v1) namespace: default
8:49:02PM: ---- waiting on 1 changes [2/3 done] ----
8:49:12PM: ongoing: reconcile deployment/kube-vip-cluster (apps/v1) namespace: default
8:49:12PM:  ^ Waiting for 2 unavailable replicas
8:49:12PM:  L ok: waiting on replicaset/kube-vip-cluster-76c6cfdd77 (apps/v1) namespace: default
8:49:12PM:  L ongoing: waiting on pod/kube-vip-cluster-76c6cfdd77-mrxnb (v1) namespace: default
8:49:12PM:     ^ Pending: Unschedulable (message: 0/4 nodes are available: 1 node(s) didn't match pod affinity/anti-affinity, 3 node(s) had taints that the pod didn't tolerate.)
8:49:12PM:  L ongoing: waiting on pod/kube-vip-cluster-76c6cfdd77-8cvvr (v1) namespace: default
8:49:12PM:     ^ Pending: Unschedulable (message: 0/4 nodes are available: 1 node(s) didn't match pod affinity/anti-affinity, 3 node(s) had taints that the pod didn't tolerate.)
8:49:12PM:  L ok: waiting on pod/kube-vip-cluster-76c6cfdd77-5mkpb (v1) namespace: default
8:50:03PM: ---- waiting on 1 changes [2/3 done] ----

[thebsdbox]

Man alive.. some tooling is too clever for it's own good!

I'd kubectl scale --replicas=1 deploy/kube-vip-cluster and if you want to submit a PR to change the manifest to 1 that's fine or I can do it and update docs later on.

[thebsdbox]

Docs updated to reflect the usage of deployment with kubectl or kapps

-> https://kube-vip.io/kubernetes/