Describe the bug
In toolkit, GKE node pool upgrade settings are hardcoded:
|
upgrade_settings { |
|
strategy = "SURGE" |
|
max_surge = 0 |
|
max_unavailable = 1 |
|
} |
This stops us from efficiently upgrading nodes in-place. Without change, each node upgrade can take up 9+ minutes, which makes maintaining big node pools unrealistic.
Steps to reproduce
Steps to reproduce the behavior:
- Trigger in-place node pool upgrade
Expected behavior
You have an option to make sure multiple nodes are made unavailable time to minimize the downtime.
Actual behavior
You don't have any option to upgrade more than one node at a time.
Describe the bug
In toolkit, GKE node pool upgrade settings are hardcoded:
cluster-toolkit/modules/compute/gke-node-pool/main.tf
Lines 62 to 66 in 26fafe0
This stops us from efficiently upgrading nodes in-place. Without change, each node upgrade can take up 9+ minutes, which makes maintaining big node pools unrealistic.
Steps to reproduce
Steps to reproduce the behavior:
Expected behavior
You have an option to make sure multiple nodes are made unavailable time to minimize the downtime.
Actual behavior
You don't have any option to upgrade more than one node at a time.