The resource usage spikes up when one of the master in a HA cluster goes down
especially on a large and dense cluster due to unintentional conditions like
power, network, infrastructure failures or intentional conditions like cluster
shutdown to save costs. This commit adds information around the conditions for
customers and users to plan their master node requirements accordingly to avoid
cascading failures.
The goal is to have underutilized nodes with space for workloads to be
squeezed while nodes are being rebooted. The typo is from 7eb9504406
(Recommendation around master node sizing to handle upgrades,
2021-05-06, #32230).
This commit adds recommendation for master node sizing around OLM
memory usage to help users and customers plan their environment
accordingly. It depemds on the number of namespaces and user
installed operators that OLM is managing on the cluster.
This commit adds information around the load on the cluster during the
control plane testing to help users and customers plan their environment
accordingly.
This commit:
- Adds a recommendation around master node sizing to handle upgrades and
failures on large and dense clusters as the resource usage is expected
to increase in such situations.
- Adds a note that the data points are only valid for OpenShift clusters
using OpenShiftSDN as the network plugin.
This commit adds information to the Scalabilty and Performance doc about the
size of master nodes based on the data collected during the control plane
density test runs at different node scale. This will help customers understand
the master node requirements depending on the number of nodes in the cluster.
