diff --git a/docs/deployment/schedulers/k3s.md b/docs/deployment/schedulers/k3s.md index 975c51061..baaebf46a 100644 --- a/docs/deployment/schedulers/k3s.md +++ b/docs/deployment/schedulers/k3s.md @@ -444,7 +444,7 @@ To enable autoscaling, use the `app.json` `formation.$PROCESS_TYPE.autoscaling` - `cooldown_seconds`: (default: 300) The number of seconds to wait in between each scaling event - `triggers`: A list of autoscaling triggers. -Autoscaling triggers are passed as is to Keda, and should match the configuration keda uses for a given [scaler](https://keda.sh/docs/2.13/scalers/). Below is an example for [datadog](https://keda.sh/docs/2.13/scalers/datadog/#example-2---driving-scale-directly): +Autoscaling triggers are passed as is to Keda, and should match the configuration keda uses for a given [scaler](https://keda.sh/docs/2.19/scalers/). Below is an example for [datadog](https://keda.sh/docs/2.19/scalers/datadog/#example-2---driving-scale-directly): ```json { @@ -742,6 +742,7 @@ Next, run the `scheduler-k3s:ensure-charts` command with the `vector` chart to f ```shell dokku scheduler-k3s:ensure-charts --charts vector ``` + Please see the [vector logs documentation](/docs/deployment/logs.md#configuring-a-log-sink) for more information on specifying vector sinks. ### Supported Resource Management Properties diff --git a/plugins/scheduler-k3s/scheduler_k3s.go b/plugins/scheduler-k3s/scheduler_k3s.go index 63d9268cf..81dff53a7 100644 --- a/plugins/scheduler-k3s/scheduler_k3s.go +++ b/plugins/scheduler-k3s/scheduler_k3s.go @@ -124,7 +124,7 @@ var HelmCharts = []HelmChart{ Namespace: "keda", ReleaseName: "keda", RepoURL: "https://kedacore.github.io/charts", - Version: "2.16.0", + Version: "2.19.0", }, { ChartPath: "keda-add-ons-http", @@ -132,7 +132,7 @@ var HelmCharts = []HelmChart{ Namespace: "keda", ReleaseName: "keda-add-ons-http", RepoURL: "https://kedacore.github.io/charts", - Version: "0.8.0", + Version: "0.12.2", }, { ChartPath: "vector",