Skip to content

Commit

Permalink
helm: flip podPriorityClassNodeCritical to true.
Browse files Browse the repository at this point in the history 
Set `podPriorityClassNodeCritical` to true by default, adding
plugins to the `system-node-critical` priority class. This
should mitigate the potential risk of a plugin getting evicted
under heavy system load. It should also ensure that during
autoscaling enough new nodes are brought up to leave room for
plugins on each new node.

Signed-off-by: Krisztian Litkey <[email protected]>
  • Loading branch information
@klihub
klihub committed Apr 9, 2024
1 parent d810e82 commit 85d2c7a
Show file tree
Hide file tree
Showing 12 changed files with 42 additions and 24 deletions.
2 changes: 1 addition & 1 deletion deployment/helm/balloons/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -107,4 +107,4 @@ customize with their own values, along with the default values.
| `tolerations` | [] | specify taint toleration key, operator and effect |
| `affinity` | [] | specify node affinity |
| `nodeSelector` | [] | specify node selector labels |
| `podPriorityClassNodeCritical` | false | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
| `podPriorityClassNodeCritical` | true | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
9 changes: 6 additions & 3 deletions deployment/helm/balloons/values.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -88,6 +88,9 @@ nodeSelector: []
# kubernetes.io/disk: "ssd"

# NRI plugins should be considered as part of the container runtime.
# Therefore, adding the system-node-critical priority class to the DaemonSet,
# could mitigate potential risk in a running system under load.
podPriorityClassNodeCritical: false
# By default we make them part of the system-node-critical priority
# class. This should mitigate the potential risk of a plugin getting
# evicted under heavy system load. It should also ensure that during
# autoscaling enough new nodes are brought up to leave room for the
# plugin on each new node.
podPriorityClassNodeCritical: true
2 changes: 1 addition & 1 deletion deployment/helm/memory-qos/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -103,4 +103,4 @@ customize with their own values, along with the default values.
| `tolerations` | [] | specify taint toleration key, operator and effect |
| `affinity` | [] | specify node affinity |
| `nodeSelector` | [] | specify node selector labels |
| `podPriorityClassNodeCritical` | false | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
| `podPriorityClassNodeCritical` | true | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
9 changes: 6 additions & 3 deletions deployment/helm/memory-qos/values.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,9 @@ nodeSelector: []
# kubernetes.io/disk: "ssd"

# NRI plugins should be considered as part of the container runtime.
# Therefore, adding the system-node-critical priority class to the DaemonSet,
# could mitigate potential risk in a running system under load.
podPriorityClassNodeCritical: false
# By default we make them part of the system-node-critical priority
# class. This should mitigate the potential risk of a plugin getting
# evicted under heavy system load. It should also ensure that during
# autoscaling enough new nodes are brought up to leave room for the
# plugin on each new node.
podPriorityClassNodeCritical: true
2 changes: 1 addition & 1 deletion deployment/helm/memtierd/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -103,4 +103,4 @@ customize with their own values, along with the default values.
| `tolerations` | [] | specify taint toleration key, operator and effect |
| `affinity` | [] | specify node affinity |
| `nodeSelector` | [] | specify node selector labels |
| `podPriorityClassNodeCritical` | false | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
| `podPriorityClassNodeCritical` | true | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
9 changes: 6 additions & 3 deletions deployment/helm/memtierd/values.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -55,6 +55,9 @@ nodeSelector: []
# kubernetes.io/disk: "ssd"

# NRI plugins should be considered as part of the container runtime.
# Therefore, adding the system-node-critical priority class to the DaemonSet,
# could mitigate potential risk in a running system under load.
podPriorityClassNodeCritical: false
# By default we make them part of the system-node-critical priority
# class. This should mitigate the potential risk of a plugin getting
# evicted under heavy system load. It should also ensure that during
# autoscaling enough new nodes are brought up to leave room for the
# plugin on each new node.
podPriorityClassNodeCritical: true
2 changes: 1 addition & 1 deletion deployment/helm/sgx-epc/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -103,4 +103,4 @@ customize with their own values, along with the default values.
| `tolerations` | [] | specify taint toleration key, operator and effect |
| `affinity` | [] | specify node affinity |
| `nodeSelector` | [] | specify node selector labels |
| `podPriorityClassNodeCritical` | false | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
| `podPriorityClassNodeCritical` | true | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
9 changes: 6 additions & 3 deletions deployment/helm/sgx-epc/values.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,9 @@ nodeSelector: []
# kubernetes.io/disk: "ssd"

# NRI plugins should be considered as part of the container runtime.
# Therefore, adding the system-node-critical priority class to the DaemonSet,
# could mitigate potential risk in a running system under load.
podPriorityClassNodeCritical: false
# By default we make them part of the system-node-critical priority
# class. This should mitigate the potential risk of a plugin getting
# evicted under heavy system load. It should also ensure that during
# autoscaling enough new nodes are brought up to leave room for the
# plugin on each new node.
podPriorityClassNodeCritical: true
2 changes: 1 addition & 1 deletion deployment/helm/template/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -105,4 +105,4 @@ customize with their own values, along with the default values.
| `initImage.tag` | unstable | init container image tag |
| `initImage.pullPolicy` | Always | init container image pull policy |
| `tolerations` | [] | specify taint toleration key, operator and effect |
| `podPriorityClassNodeCritical` | false | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
| `podPriorityClassNodeCritical` | true | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
9 changes: 6 additions & 3 deletions deployment/helm/template/values.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -55,6 +55,9 @@ tolerations: []
# effect: "NoSchedule"

# NRI plugins should be considered as part of the container runtime.
# Therefore, adding the system-node-critical priority class to the DaemonSet,
# could mitigate potential risk in a running system under load.
podPriorityClassNodeCritical: false
# By default we make them part of the system-node-critical priority
# class. This should mitigate the potential risk of a plugin getting
# evicted under heavy system load. It should also ensure that during
# autoscaling enough new nodes are brought up to leave room for the
# plugin on each new node.
podPriorityClassNodeCritical: true
2 changes: 1 addition & 1 deletion deployment/helm/topology-aware/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -108,4 +108,4 @@ customize with their own values, along with the default values.
| `tolerations` | [] | specify taint toleration key, operator and effect |
| `affinity` | [] | specify node affinity |
| `nodeSelector` | [] | specify node selector labels |
| `podPriorityClassNodeCritical` | false | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
| `podPriorityClassNodeCritical` | true | enable [marking Pod as node critical](https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/#marking-pod-as-critical) |
9 changes: 6 additions & 3 deletions deployment/helm/topology-aware/values.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -76,6 +76,9 @@ nodeSelector: []
# kubernetes.io/disk: "ssd"

# NRI plugins should be considered as part of the container runtime.
# Therefore, adding the system-node-critical priority class to the DaemonSet,
# could mitigate potential risk in a running system under load.
podPriorityClassNodeCritical: false
# By default we make them part of the system-node-critical priority
# class. This should mitigate the potential risk of a plugin getting
# evicted under heavy system load. It should also ensure that during
# autoscaling enough new nodes are brought up to leave room for the
# plugin on each new node.
podPriorityClassNodeCritical: true

0 comments on commit 85d2c7a

Please sign in to comment.