In some cases it might be usefull to limit how many evictions per a
domain can be performed. To avoid burning the whole per descheduling
cycle budget. Limiting the number of evictions per node is a
prerequisite for evicting pods whose usage can't be easily subtracted
from overall node resource usage to predict the final usage. E.g. when a
pod is evicted due to high PSI pressure which takes into account many
factors which can be fully captured by the current predictive resource
model.
This commit adds a sample plugin implementation as follow:
This directory provides an example plugin for the Kubernetes Descheduler,
demonstrating how to evict pods based on custom criteria. The plugin targets
pods based on:
* **Name Regex:** Pods matching a specified regular expression.
* **Age:** Pods older than a defined duration.
* **Namespace:** Pods within or outside a given list of namespaces (inclusion
or exclusion).
To incorporate this plugin into your Descheduler build, you must register it
within the Descheduler's plugin registry. Follow these steps:
1. **Register the Plugin:**
* Modify the `pkg/descheduler/setupplugins.go` file.
* Add the following registration line to the end of the
`RegisterDefaultPlugins()` function:
```go
pluginregistry.Register(
example.PluginName,
example.New,
&example.Example{},
&example.ExampleArgs{},
example.ValidateExampleArgs,
example.SetDefaults_Example,
registry,
)
```
2. **Generate Code:**
* If you modify the plugin's code, execute `make gen` before rebuilding the
Descheduler. This ensures generated code is up-to-date.
3. **Rebuild the Descheduler:**
* Build the descheduler with your changes.
Configure the plugin's behavior using the Descheduler's policy configuration.
Here's an example:
```yaml
apiVersion: descheduler/v1alpha2
kind: DeschedulerPolicy
profiles:
- name: LifecycleAndUtilization
plugins:
deschedule:
enabled:
- Example
pluginConfig:
- name: Example
args:
regex: ^descheduler-test.*$
maxAge: 3m
namespaces:
include:
- default
```
- `regex: ^descheduler-test.*$`: Evicts pods whose names match the regular
expression `^descheduler-test.*$`.
- `maxAge: 3m`: Evicts pods older than 3 minutes.
- `namespaces.include: - default`: Evicts pods within the default namespace.
This configuration will cause the plugin to evict pods that meet all three
criteria: matching the `regex`, exceeding the `maxAge`, and residing in the
specified namespace.
When the feature is enabled each pod with descheduler.alpha.kubernetes.io/request-evict-only
annotation will have the eviction API error examined for a specific
error code/reason and message. If matched eviction of such a pod will be interpreted
as initiation of an eviction in background.
* add ignoreNonPDBPods option
* take2
* add test
* poddisruptionbudgets are now used by defaultevictor plugin
* add poddisruptionbudgets to rbac
* review comments
* don't use GetPodPodDisruptionBudgets
* review comment, don't hide error
* skip eviction when pod creation time is below minPodAge threshold setting
In the default initialization phase of the descheduler, add a new
constraint to not evict pods that creation time is below minPodAge
threshold.
Added value:
- Avoid crazy pod movement when the autoscaler scales up and down.
- Avoid evicting pods when they are warming up.
- Decreases the overall cost of eviction as no pod will be evicted
before doing significant amount of work.
- Guard against scheduling. Descheduling loops in situations where
the descheduler has a different node fit logic from scheduler,
like not considering topology spread constraints.
* Use *time.Duration instead of uint for MinPodAge type
* Remove '(in minutes)' from default evictor configuration table
* make fmt
* Add explicit name for Duration field
* Use Duration.String()
Currently, all the plugins are run in a sequence.
No plugin executes evictions in parallel within.
Yet, there's no guarantee a future plugin (e.g. a custom one)
will not attemp to evict pods in parallel.
