Hi,
I'm trying the kubernetes elasticsearch operator, and I have defined an EDS with the following scaling attributes:
scaling:
enabled: true
minReplicas: 1
maxReplicas: 4
minIndexReplicas: 0
maxIndexReplicas: 2
minShardsPerNode: 10
maxShardsPerNode: 20
scaleUpCPUBoundary: 75
scaleUpCooldownSeconds: 240
scaleUpThresholdDurationSeconds: 180
scaleDownCPUBoundary: 50
scaleDownCooldownSeconds: 120
scaleDownThresholdDurationSeconds: 30
diskUsagePercentScaledownWatermark: 75
The cluster starts with one data node, and if I create an index like this:
curl -XPUT localhost:9200/demo-index -HContent-type:application/json -d '{"settings": {"index": { "number_of_shards":11, "number_of_replicas":2}}}'
The operator creates another node. So far so good. But if I delete the index (thus removing the need for two data nodes), the operator does not scale down automatically. If I modify minimally the EDS yaml and apply it to kubernetes, the operator immediately scales down, so probably I'm missing some configuration here. I thought that the setting was "scaleDownThresholDurationSeconds" but it does not seem to work.
What am I missing here? Is it not supposed to scale down automatically in this case?