Feature Overview:

Too big of a node, too many nodes, etc. Similar to ROS for RHEL.

Scope and what’s possible TBD, needs conversation with Kruize and AppLearner teams.

• Define default utilization level for warning (e. g. low = 20% and high = 80%)
• Allow users to define their own default utilization levels warning
• Allow users to define their on utilization levels warning for the nodes of this or that cluster
• Allow users to define their on utilization levels warning for control plane nodes vs worker plane nodes
• Allow users to define specific utilization levels warning per instance size (e. g. 70% of small worker node size might be critical but 90% of a large worker node might be OK)
• Recommend "according to X time observation, we think you can downsize your cluster from X nodes to Y nodes" (where X > Y) or viceversa
• Look at unscheduled jobs
• Look at average, peak, valley, time over high warning level, time under low warning level, etc
• Let users define what instance types they can use. Typically organizations standardize (or get better deals) on some instance types.
• Let users define on what cloud regions they can operate and how many they can have at minimum/maximum. Typically organizations standardize on some regions because of compliance, proximity, etc.

Docs:

Yes

SMEs and Stakeholders: