Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

docs: Adjust wording and examples #4896

Merged
merged 4 commits into from
Dec 6, 2024
Merged

docs: Adjust wording and examples #4896

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
93cd245
Adjust wording and examples
kreeuwijk Dec 6, 2024
fabcc1a
ci: auto-formatting prettier issues
kreeuwijk Dec 6, 2024
27c44ef
docs: update patch
lennessyy Dec 6, 2024
b1b475a
docs: add edits
lennessyy Dec 6, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
45 changes: 25 additions & 20 deletions docs/docs-content/clusters/edge/cluster-management/skip-draining.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
---
sidebar_label: "Skip Node Draining During Upgrades"
title: "Skip Node Draining During Upgrades"
description: "Learn how to skip node draining during cluster upgrades. "
sidebar_label: "Skip Node Draining
title: "Skip Node Draining
description: "Learn how to skip node draining during cluster upgrades and cluster repaves."
icon: ""
sidebar_position: 40
hide_table_of_contents: false
Expand All @@ -15,8 +15,11 @@ minimize application downtime. For more information what changes will cause rebo

However, the benefits of draining a node in a single-node cluster are minimal because there are no other nodes to
schedule the workloads onto. In addition, if system-critical workloads are drained, the cluster may get stuck in an
unmanageable state. You can configure your cluster to skip node draining to avoid such outcomes. These configurations
will apply to both appliance mode and [agent mode](../../../deployment-modes/agent-mode/agent-mode.md) deployments.
unmanageable state.

By default, Palette will only drain nodes in multi-node Edge clusters. You can configure draining behavior via the OS
layer of the cluster profile. These configurations will apply to both appliance mode and
[agent mode](../../../deployment-modes/agent-mode/agent-mode.md) deployments.

## Prerequisites

Expand All @@ -28,38 +31,40 @@ will apply to both appliance mode and [agent mode](../../../deployment-modes/age
[Edge Host Registration](../site-deployment/site-installation/edge-host-registration.md) and
[Deployment](../site-deployment/site-deployment.md).

## Skip Node Draining During Upgrades
- Palette version 4.5.b or higher, with Palette agent version 4.5.11 or higher.

## Skip Node Draining During Upgrades or Repaves

1. Log in to [Palette](https://console.spectrocloud.com).

2. From the left **Main Menu**, click **Profiles**.

3. Select the cluster profile you use to provision the clusters for which you want to skip node draining during
upgrades.
3. Select the cluster profile you use to provision the clusters for which you want to skip node draining during upgrades
or repaves.

4. Select the operating system layer of your profile, and add the following lines.
4. Select the BYOS Edge OS layer of your profile, and configure the `pack.drain` section.

```yaml {2,3}
pack:
drainPods: false
podSelector: upgrade.cattle.io/plan!=control-plan,upgrade.cattle.io/plan!=worker-plan,app!=spectro,app!=spectro-proxy,app!=palette-webhook
drain:
drainPods: false
```

The following table provides a brief description of the parameters.

| Parameter | Description | Default |
| ------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ------- |
| `drainPods` | Controls the node drain behavior during a cluster upgrade. `true` means nodes will be drained. `false` means nodes will not be drained. `auto` means nodes will only be drained for multi-node clusters, while single-node clusters will not drain nodes during an upgrade. | `auto` |
| `podSelector` | If `drainPods` is set to `true` for either single-node or multi-node clusters, or set to `auto` for multi-node clusters, only pods matching the selectors will be drained. <br /> <br /> Pods with any of the following labels are always protected from draining: `upgrade.cattle.io/plan=control-plan`,`upgrade.cattle.io/plan=worker-plan`,`app=spectro`,`app=spectro-proxy` ,`app=palette-webhook` | None |
| Parameter | Description | Default |
| ------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------- |
| `drainPods` | Controls the node drain behavior during a cluster upgrade. `true` means nodes will always be drained. `false` means nodes will never be drained. `auto` means nodes will only be drained for multi-node clusters, while single-node clusters will not drain nodes during an upgrade. | `auto` |
| `podSelector` | If `drainPods` is set to `true` for either single-node or multi-node clusters, or set to `auto` for multi-node clusters, only pods matching the selectors will be drained. You can use positive or negative matches. For example, `upgrade.cattle.io/plan!=control-plan` means pods with the `upgrade.cattle.io/plan=control-plan` label are not drained, and `app=web` means pods with the label `app=web` are drained. <br /> <br /> Pods with any of the following labels are always protected from draining: `upgrade.cattle.io/plan=control-plan`,`upgrade.cattle.io/plan=worker-plan`,`app=spectro`,`app=spectro-proxy` ,`app=palette-webhook` | None |

:::warning

In single-node clusters, disabling node draining means normal workloads and the upgrade process happen in parallel.
This will increase memory usage, and may cause your node to become unresponsive if your host is memory-constrained.

In such cases, you may set the `pack.drainPods` parameter to `true`, and set `pack.disableEviction` to `true`. This
will prevent the drain process from hanging indefinitely due to `PodDisruptionBudget` constraints, while the default
`podSelector` will protect most critical pods.
In such cases, you may set the `pack.drain.drainPods` parameter to `true`, and set `pack.drain.disableEviction` to
`true`. This will prevent the drain process from hanging indefinitely due to `PodDisruptionBudget` constraints, while
the default `podSelector` will protect most critical system pods.

:::

Expand All @@ -79,5 +84,5 @@ will apply to both appliance mode and [agent mode](../../../deployment-modes/age
[Edge Cluster Upgrade Behavior](../cluster-management/upgrade-behavior.md).

2. After the upgrade is completed, use `kubectl logs` to check on a node for which you skipped pod draining. Confirm
that no messages that look like `Evicting pod <pod-name> as part of upgrade plan` are displayed. The absence of
such messages means that the pods were not drained during the upgrade.
that no messages that look like `Evicting pod <pod-name> as part of upgrade plan` are displayed. The absence of such
messages means that the pods were not drained during the upgrade.
Loading