mirror of
https://github.com/ubicloud/ubicloud.git
synced 2025-10-05 06:12:09 +08:00
c44348131d
To determine the total effective vCPU usage for each Kubernetes cluster,
sum the vCPUs of all control plane nodes and all worker nodes via node
pools.
Initially, I implemented something similar to the PostgreSQL vCPU usage
calculation to keep the entire logic within the database.
(kubernetes_clusters_dataset.association_join(nodes: :vm).sum(:vcpus) || 0) + (kubernetes_clusters_dataset.association_join(nodepools: {nodes: :vm}).sum(:vcpus) || 0)
The issue with summing vCPUs of all control plane and worker nodes is
that it assumes all nodes are already created. Kubernetes clusters
differ from PostgreSQL resources: node pools are created at the next
labels of the nexus, not immediately in the assemble.
If a customer creates many clusters at once, the vCPU usage will show as
0 until the strand runs and actually provisions the nodes. To handle
this, I decided to base part of the calculation on the planned target
node sizes (target_node_sizes) instead of relying solely on existing
nodes.
|
||
|---|---|---|
| .. | ||
| clec | ||
| lib | ||
| model | ||
| prog | ||
| routes | ||
| scheduling | ||
| serializers | ||
| cli_config.ru | ||
| cli_spec.rb | ||
| config_spec.rb | ||
| coverage_helper.rb | ||
| db_spec.rb | ||
| monitor_smoke_test.rb | ||
| resource_methods_spec.rb | ||
| respirate_smoke_test.rb | ||
| ruby_sdk_spec.rb | ||
| spec_helper.rb | ||
| thawed_mock.rb | ||
| ubid_spec.rb | ||