More scaled down than scaled up

[dbluxo]

Hi,

first of all, thank you for your project!

I've tried kube-downscaler in a nearly empty cluster. Here are the logs:

2019-03-12 12:52:11,615 INFO: Downscaler v0.9 started with debug=False, default_downtime=never, default_uptime=Mon-Fri 07:30-20:00 Europe/Berlin, dry_run=False, exclude_deployments=kube-downscaler,downscaler, exclude_namespaces=kube-system, exclude_statefulsets=, grace_period=900, interval=60, kind=['deployment', 'deployment', 'statefulset'], namespace=None, once=False
2019-03-12 19:00:23,032 INFO: Scaling down Deployment authelia/authelia-app from 1 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,045 INFO: Scaling down Deployment authelia/authelia-redis-slave from 2 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,060 INFO: Scaling down Deployment cattle-system/cattle-cluster-agent from 1 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,080 INFO: Scaling down Deployment logging/elasticsearch-client from 2 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,103 INFO: Scaling down Deployment logging/elasticsearch-exporter from 1 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,124 INFO: Scaling down Deployment logging/kibana from 1 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,144 INFO: Scaling down Deployment logging/laas-metricbeat from 1 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,161 INFO: Scaling down Deployment monitoring/kube-state-metrics from 1 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,191 INFO: Scaling down StatefulSet authelia/authelia-redis-master from 1 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,212 INFO: Scaling down StatefulSet authelia/mongo from 3 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,231 INFO: Scaling down StatefulSet logging/elasticsearch-data from 3 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,253 INFO: Scaling down StatefulSet logging/elasticsearch-master from 3 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,271 INFO: Scaling down StatefulSet monitoring/alertmanager from 3 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,290 INFO: Scaling down StatefulSet monitoring/grafana from 1 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-12 19:00:23,310 INFO: Scaling down StatefulSet monitoring/prometheus from 2 to 0 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-13 06:30:14,204 INFO: Scaling up Deployment authelia/authelia-app from 0 to 1 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-13 06:30:14,216 INFO: Scaling up Deployment authelia/authelia-redis-slave from 0 to 2 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-13 06:30:14,228 INFO: Scaling up Deployment cattle-system/cattle-cluster-agent from 0 to 1 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-13 06:30:14,239 INFO: Scaling up Deployment logging/elasticsearch-client from 0 to 2 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-13 06:30:14,252 INFO: Scaling up Deployment logging/elasticsearch-exporter from 0 to 1 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-13 06:30:14,268 INFO: Scaling up Deployment logging/kibana from 0 to 1 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-13 06:30:14,286 INFO: Scaling up Deployment logging/laas-metricbeat from 0 to 1 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-13 06:30:14,299 INFO: Scaling up Deployment monitoring/kube-state-metrics from 0 to 1 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-13 06:30:14,345 INFO: Scaling up StatefulSet monitoring/alertmanager from 0 to 3 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-13 06:30:14,358 INFO: Scaling up StatefulSet monitoring/grafana from 0 to 1 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)
2019-03-13 06:30:14,372 INFO: Scaling up StatefulSet monitoring/prometheus from 0 to 2 replicas (uptime: Mon-Fri 07:30-20:00 Europe/Berlin, downtime: never)

As you can see: 15 x "Scaling down", but only 11 x "Scaling up". I have tried it several times, at each try this statefulsets were not scaled up:

• authelia/authelia-redis-master
• authelia/mongo
• logging/elasticsearch-data
• logging/elasticsearch-master

We use kube-downscaler version 0.9. (Kubernetes v1.12.6)

Any idea why?

[hjacobs]

This might be related to #21

[dbluxo]

Yes, I can confirm that it only affects StatefulSets, which hasn't previous annotations and adding annotation downscaler/original-replicas fails. I think it doesn't affect Deployments, because the annotation deployment.kubernetes.io/revision exists automatically and adding further annotation downscaler/original-replicas works.

[hjacobs]

Should be fixed by #42 and released in v0.11: https://github.com/hjacobs/kube-downscaler/releases/tag/0.11