Node label not applied when node running kubeIP has its IP address forcibly changed

[doitintl]

Node label not applied when node running kubeIP has its IP address forcibly changed

[spark2ignite]

@yob Under the deploy/kubeip-configmap.yaml there is a KUBEIP_FORCEASSIGNMENT param which is set to true by default.

If you don't want your existing nodes to change the existing external IP, modify it to false before deploying kubeIP. When the KUBEIP_FORCEASSIGNMENT is set to false, kubeIP will assign static external IPs from the pool only to the new nodes in the node-pool.

Now, regarding the read: connection reset by peer, we will fix it ASAP. Thanks for reporting.

[yob]

Cheers. I considered that, but I was hoping to keep KUBEIP_FORCEASSIGNMENT on, yet still get the labels set.

It's a bit of a tricky timing issue I guess, and maybe it's not worth the effort to fix the edge case. Would it be possible to have the ticker set the labels if they're missing? We wouldn't need the labels straight away, having the cluster to converge on the final config after a few minutes would be fine (that's more or less the way everything works in kubernetes, so we're used to it)

[avivl]

@yob I'll supply a solution either today or early next week

[avivl]

@yob Done.
On the ticker, I'm looking for nodes that got reserved IP but do not have a tag and then I tag them

[yob]

Thanks!