powerfulseal/powerfulseal

Node dissapears if restarted on a GKE cluster

Open

#133 opened on 2019年8月23日

GitHub で見る
 (1 comment) (0 reactions) (0 assignees)Python (1,890 stars) (170 forks)batch import
buggood first issuehelp wanted

説明

Describe the bug When running Powerfulseal on a GKE cluster, if you restart a node (using stop/start) it disappears from the node list.

To Reproduce Steps to reproduce the behavior:

  1. Set up a GKE cluster with the default configuration
  2. Launch Powerfulseal on interactive mode pointing to the cluster (as it's easier for testing)
  3. Do: nodes
  4. Do: stop
  5. Wait for the node to stop
  6. Do: sync
  7. Do: start
  8. Wait for the node to start (check GCP Web Panel)
  9. Do: sync
  10. Node is missing!

Expected behavior I should be able to see my restarted nodes!

What's happening The external IP on GCP disappears when a node is stopped, and a new one is assigned dynamically when restarted. But on GKE the internal IP changes on restart too! So when Powerfulseal tries to match the IP's from the inventory to any cloud node... They are all missing.

When running a non-GKE cluster on top of standard GCP Compute nodes, the internal IP is always the same, so using kubernetes-inventory or ansible-like inventory works as expected.

コントリビューターガイド