The bbb_scaler scales big blue button intances in the ionos cloud using metrics of node exporter, big blue button api and prometheus.
This project is involved as autoscaler in the terraform setup of schul cloud.
The target machines need metric interfaces of big blue button, node exporter and a registration in prometheus. The node exporter metrics have to be exposed on port 9100.
This solution was built to use the ionos api v5.
- Get all machines in the ionos data center
- Filter all machines
- Get only big blue button instances
- Get only running instances
- Retrieve workload
- Get CPU Utilization via Graphana and Prometheus
- Get Memory Utilization via Node Exporter
- Check if there are still active meetings
- Start Scaling
- If the average workload of all machines is higher than the allowed cpu or memory utilization, start a new instance
- If the workload of an instance has the minimum cpu and memory settings and is still not utilized, shut it down
- If the memory workload of an instance is lower than the down scaling threshold MIN_ALLOWED_MEMORY_WORKLOAD and the memory is higher than DEFAULT_WORKER_MEMORY, decrease memory by 1024 bytes
- If the cpu workload of an instance is lower than the down scaling threshold MIN_ALLOWED_CPU_WORKLOAD and the cores are higher than DEFAULT_WORKER_MEMORY, decrease cores by 1 unit
- If the memory workload of an instance is higher than the scaling threshold MAX_ALLOWED_MEMORY_WORKLOAD and the memory is lower than or equal MAX_WORKER_MEMORY - 1024, increase memory by 1024 bytes
- If the cpu workload of an instance is higher than the scaling threshold MAX_ALLOWED_CPU_WORKLOAD and the cores + 1 are lower than or equal MAX_WORKER_CPU - 1, increase cores by 1 unit
- Wait for WAITING_TIME miliseconds
The autoscaler can be setup with environment variables.
| Name | Default Value | Description |
|---|---|---|
| MINIMUM_ACTIVE_MACHINES | 2 | Amount of big blue button instances that should always active |
| MAX_ALLOWED_MEMORY_WORKLOAD | 0.35 | Maximum memory utilization until up scaling starts |
| MAX_ALLOWED_CPU_WORKLOAD | 0.35 | Maximum cpu utilization until up scaling starts |
| MIN_ALLOWED_MEMORY_WORKLOAD | 0.15 | Minimum memory utilization until down scaling starts |
| MIN_ALLOWED_CPU_WORKLOAD | 0.05 | Minimum cpu utilization until down scaling starts |
| MAX_WORKER_MEMORY | 16384 | Ceiling limit of memory scaling |
| DEFAULT_WORKER_MEMORY | 8192 | Floor limit of memory scaling |
| MAX_WORKER_CPU | 4 | Ceiling limit of cpu scaling |
| DEFAULT_WORKER_CPU | 2 | Floor limit of cpu scaling |
| WAITING_TIME | 300000 | Time in miliseconds until next scaling cyclus starts |
Secrets can be also applied using environment variables
| Name | Description |
|---|---|
| IONOS_USER | User name for ionos cloud |
| IONOS_PASS | Password for ionos cloud user |
| IONOS_DATACENTER | UUID of the data center in ionos where are the big blue button instaces located |
| BBB_PASS | Key to access metric values of the big blue button instaces |
| GRAPHANA_PASS | API Key to call prometheus via graphana |
| NE_BASIC_AUTH_USER | Node exporter basic authentication user |
| NE_BASIC_AUTH_PASS | Node exporter basic authentication password |
Use the Dockerfile to build the application. You don't need additional tools because the file already contains all build utils.
Setup all secrets in the compose file using the environment variable section