Systems with only three worker nodes (typically Test and Development Systems (TDS)) will encounter pod scheduling issues when worker nodes are taken out of the Kubernetes cluster to be upgraded.
For systems with only three worker nodes, execute the following script to reduce the CPU request for some services with high CPU requests, in order to allow critical upgrade-related services to be successfully scheduled on only two worker nodes:
ncn-mw# /usr/share/doc/csm/upgrade/1.0.1/scripts/k8s/tds_lower_cpu_requests.sh
Note that some services with these lower CPU request may encounter CPU throttling (see Determine if Pods are Hitting Resource Limits).
If needed, the top portion of the script (see below) can be edited and re-run to further adjust these CPU requests. Note that commenting out any of the lines below will indicate the script should not
adjust the CPU resource for that service. Also see the Kubernetes/Compute Resources/Namespace (Pods) Grafana page for a historical view of a given pod’s CPU utilization for this specific system. See
Access System Management Health Services.
spire_postgres_new_limit=1000m
elasticsearch_master_new_cpu_request=1500m
cluster_kafka_new_cpu_request=1
sma_grafana_new_cpu_request=100m
sma_kibana_new_cpu_request=100m
cluster_zookeeper_new_cpu_request=100m
cray_smd_new_cpu_request=1
cray_smd_postgres_new_cpu_request=500m
sma_postgres_cluster_new_cpu_request=500m
cray_capmc_new_cpu_request=500m
nexus_new_cpu_request=2
cray_metallb_speaker_new_cpu_request=1
The script will output the current value of the request along with the new value. It is recommended to capture and save the output from this script, and store it external to the cluster. This record of the value changes is helpful in the event that rolling back to the original value is desired.
Example output:
Patching cray-capmc deployment with new cpu request of 500m (from 2100m)
deployment.apps/cray-capmc patched
Waiting for deployment spec update to be observed...
Waiting for deployment "cray-capmc" rollout to finish: 0 out of 3 new replicas have been updated...
Waiting for deployment "cray-capmc" rollout to finish: 1 out of 3 new replicas have been updated...
Waiting for deployment "cray-capmc" rollout to finish: 1 out of 3 new replicas have been updated...
Waiting for deployment "cray-capmc" rollout to finish: 1 out of 3 new replicas have been updated...
Waiting for deployment "cray-capmc" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "cray-capmc" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "cray-capmc" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "cray-capmc" rollout to finish: 1 old replicas are pending termination...
Waiting for deployment "cray-capmc" rollout to finish: 1 old replicas are pending termination...
deployment "cray-capmc" successfully rolled out