github/kubernetes
1
0
Fork 0
mirror of https://github.com/k3s-io/kubernetes.git synced 2025-07-25 04:33:26 +00:00
Code Issues Projects Releases Wiki Activity

Increase time to wait for nodes to become unready

Browse Source 
https://github.com/kubernetes/autoscaler/pull/3924 changed Cluster
Autoscaler behavior to mark nodes as unhealthy only if at least 15m
passed since node creation time.
This commit is contained in:
Daniel Kłobuszewski 2021-08-12 11:03:38 +02:00
parent 1181421112
commit 05beda5c40
1 changed files with 5 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
test/e2e/autoscaling/cluster_size_autoscaling.go
View File

@ -888,15 +888,11 @@ var _ = SIGDescribe("Cluster size autoscaling [Slow]", func() {
// If new nodes are disconnected too soon, they'll be considered not started
// instead of unready, and cluster won't be considered unhealthy.
//
// More precisely, Cluster Autoscaler compares last transition time of
// several readiness conditions to node create time. If it's within
// 2 minutes, it'll assume node is just starting and not unhealthy.
//
// Nodes become ready in less than 1 minute after being created,
// so waiting extra 2 minutes before breaking them (which triggers
// readiness condition transition) should be sufficient, while
// making no assumptions about minimal node startup time.
time.Sleep(2 * time.Minute)
// More precisely, Cluster Autoscaler will never consider a
// node to be unhealthy unless it was created more than 15m
// ago. Within that 15m window, it'll assume node is just
// starting and not unhealthy.
time.Sleep(15 * time.Minute)
ginkgo.By("Block network connectivity to some nodes to simulate unhealthy cluster")
nodesToBreakCount := int(math.Ceil(math.Max(float64(unhealthyClusterThreshold), 0.5*float64(clusterSize))))