Skip to content

Latest commit

 

History

History
217 lines (159 loc) · 7.97 KB

autotune_install.md

File metadata and controls

217 lines (159 loc) · 7.97 KB
Raw

Kruize Autotune Installation

Clone Repositories

Clone the Autotune Repository

  • Clone the kruize autotune repository onto your workstation with the following command.
$ git clone [email protected]:kruize/autotune.git

Clone the Kruize Benchmarks Repository

  • Clone the kruize benchmarks repository onto your workstation with the following command.
$ git clone [email protected]:kruize/benchmarks.git

Install the required software

Kubernetes

Autotune can be deployed to a supported Kubernetes cluster. We currently support Minikube.

Install kubectl and minikube

Install kubectl

$ curl -s https://storage.googleapis.com/kubernetes-release/release/stable.txt)/bin/linux/amd64/kubectl

$ chmod +x ./kubectl

$ sudo mv ./kubectl /usr/local/bin/kubectl

$ kubectl version --client

Minikube setup

Minikube setup with 8 CPUs and 16 GB Memory is recommended for autotune deployment. After setting up minikube, install prometheus from autotune repo with the following command

Only a specific version of minikube v1.26.0 is compatible with prometheus version v0.8.0, hence preferred to install this version on your machine by:

$ curl -LO https://github.com/kubernetes/minikube/releases/download/v1.26.1/minikube-linux-amd64

$ sudo install minikube-linux-amd64 /usr/local/bin/minikube

$ minikube start --cpus=8 --memory=16384M

$ ./scripts/prometheus_on_minikube.sh -as 

    Info: installing prometheus...
    
    Info: Checking pre requisites for prometheus...
    No resources found in monitoring namespace.
    Info: Downloading cadvisor git
    
    Info: Installing cadvisor
    namespace/cadvisor created
    serviceaccount/cadvisor created
    daemonset.apps/cadvisor created
    
    Info: Downloading prometheus git release - v0.8.0
    Info: Installing prometheus
    namespace/monitoring created
    customresourcedefinition.apiextensions.k8s.io/alertmanagerconfigs.monitoring.coreos.com created
    customresourcedefinition.apiextensions.k8s.io/alertmanagers.monitoring.coreos.com created
    Info: Waiting for all Prometheus Pods to get spawned....done
    
    Info: Waiting for prometheus-k8s-1 to come up.....
    prometheus-k8s-1                       0/2     ContainerCreating            0          70s
    prometheus-k8s-1                       0/2     Pending             0          74s
    prometheus-k8s-1                       2/2     Running   1    85s
    Info: prometheus-k8s-1 deploy succeeded: Running
    prometheus-k8s-1                       2/2     Running   1    85s

Build autotune docker image

Build autotune docker image 'autotune:test' with the following command

$ ./build.sh -i autotune:test


   Usage: ./build.sh [-d] [-v version_string] [-i autotune_docker_image]
	-d: build in dev friendly mode
	-i: build with specific autotune operator docker image name [Default - kruize/autotune_operator:<version from pom.xml>]
	-v: build as specific autotune version

After building the docker image push the image

1.Login into the docker account

Click here to create an account in docker hub Docker

$ docker login

$ docker tag autotune:test docker.io/username/autotune:test

2.Push the Docker image

$ docker push username/autotune:test

Note - You can use the 'dev friendly mode' option to quickly build the autotune docker image using the cached layers.

Deploy Autotune

Let us now deploy autotune using the docker image onto the minikube cluster

$ ./deploy.sh -c minikube -i autotune:test

  Usage: ./deploy.sh [-a] [-k url] [-c [docker|minikube|openshift]] [-i autotune docker image] [-o hpo docker image] [-n namespace] [-d configmaps-dir ] [--timeout=x, x in seconds, for docker only]
         -s = start(default), -t = terminate
         -s: Deploy autotune [Default]
         -t: Terminate autotune deployment
         -c: kubernetes cluster type. At present we support only minikube [Default - minikube]
         -i: build with specific autotune operator docker image name [Default - kruize/autotune_operator:<version from pom.xml>]
         -o: build with specific hpo docker image name [Default - kruize/hpo:0.0.2]
         -n: Namespace to which autotune is deployed [Default - monitoring namespace for cluster type minikube]
         -d: Config maps directory [Default - manifests/configmaps]
         -m: Target mode selection [autotune | crc]
         -d: Config maps directory [Default - manifests/configmaps]
        Unknown option --help
        
 
  For example,
 
  ./deploy.sh -c minikube -m autotune -i docker.io/<dockerhub username>/autotune:test
    
    Info: Checking pre requisites for minikube...
    Prometheus is installed and running.
    Create autotune namespace monitoring
    Info: One time setup - Create a service account to deploy autotune
    serviceaccount/autotune-sa created
    customresourcedefinition.apiextensions.k8s.io/autotunes.recommender.com created
    customresourcedefinition.apiextensions.k8s.io/autotuneconfigs.recommender.com created
    customresourcedefinition.apiextensions.k8s.io/autotunequeryvariables.recommender.com created
    customresourcedefinition.apiextensions.k8s.io/kruizeperformanceprofiles.recommender.com created
    clusterrole.rbac.authorization.k8s.io/autotune-cr created
    clusterrolebinding.rbac.authorization.k8s.io/autotune-crb created
    clusterrolebinding.rbac.authorization.k8s.io/autotune-prometheus-crb created
    clusterrolebinding.rbac.authorization.k8s.io/autotune-docker-crb created
    clusterrolebinding.rbac.authorization.k8s.io/autotune-scc-crb created
    autotunequeryvariable.recommender.com/minikube created
    servicemonitor.monitoring.coreos.com/autotune created
    prometheus.monitoring.coreos.com/prometheus created
    
    Creating environment variable in minikube cluster using configMap
    configmap/autotune-config created
    
    Deploying AutotuneConfig objects
    kruizelayer.recommender.com/container created
    kruizelayer.recommender.com/hotspot created
    kruizelayer.recommender.com/openj9 created
    kruizelayer.recommender.com/quarkus created
    
    Deploying Performance Profile objects
    kruizeperformanceprofile.recommender.com/resource-optimization-openshift created
    Info: Deploying autotune yaml to minikube cluster
    deployment.apps/autotune created
    service/autotune created
    Info: Waiting for autotune to come up.....
    autotune-7fd48dd988-xd6x5              0/2     ContainerCreating   0               4s
    autotune-7fd48dd988-xd6x5              0/2     ContainerCreating   0               8s
    autotune-7fd48dd988-xd6x5              0/2     ContainerCreating   0               12s
    autotune-7fd48dd988-xd6x5              0/2     ContainerCreating   0               17s
    autotune-7fd48dd988-xd6x5              0/2     ContainerCreating   0               21s
    autotune-7fd48dd988-xd6x5              0/2     ContainerCreating   0               25s
    autotune-7fd48dd988-xd6x5              0/2     ContainerCreating   0               29s
    autotune-7fd48dd988-xd6x5              0/2     ContainerCreating   0               33s
    autotune-7fd48dd988-xd6x5              0/2     ContainerCreating   0               37s
    autotune-7fd48dd988-xd6x5              0/2     ContainerCreating   0               41s
    autotune-7fd48dd988-xd6x5              0/2     ContainerCreating   0               45s
    autotune-7fd48dd988-xd6x5              2/2     Running   0               49s
    Info: autotune deploy succeeded: Running
    autotune-7fd48dd988-xd6x5              2/2     Running   0               50s
    
    Info: Access Autotune at http://192.168.49.2:30523/listKruizeTunables
    ```
    
# Demo

Please see the [demo](https://github.com/kruize/kruize-demos) repo for more details on how to use Autotune.