Introduction
Today many enterprises running Kubernetes applications are not aware of a way to monitor and optimize the resources used by those applications.
StormForge is a platform that helps users to automate resource tuning and deliver the best Kubernetes application performance at the lowest possible cost. StormForge allows users to run an experiment with a user specified number of trials. The trial runs on an application to find the optimized resource configuration required to perform certain operations. It uses machine learning algorithms to pick the combination of CPU, RAM and other resources to run a trial.
Install and configure StormForge with T4K
Follow below instructions to install and configure the StormForge with Trilio for Kubernetes and run the experiment to find the optimized resource configuration of T4K
Install and Configure StormForge redsky-controller-manager with K8s Cluster
Download and install redskyctl CLI
Copy curl -L https://app.stormforge.io/downloads/redskyctl-linux-amd64.tar.gz | tar x
sudo mv redskyctl /usr/local/bin/ 2. Authorize the linux server from where redskyctl will be evoked to run the experiment
Copy redskyctl login --force Enter the URL generated by above command into the browser to authorize the Linux server
Copy https://auth.carbonrelay.io/authorize?audience =https%3A%2F%2Fapi.carbonrelay.io%2Fv1%2F & client_id = pE3kMKdrMTdW4DOxQHesyAuFGNOWaEke & code_challenge = nUaqSbjv0grweFAMrv_Zk9KBg_8o-4AQAnwxpijJlGk & code_challenge_method = S256 & redirect_uri = http%3A%2F%2F127.0.0.1%3A8085%2F & response_type = code & scope = register%3Aclients+offline_access & state = nHGW89zTYO39loEALGTsig 3. Verify if you are connected to the server where you want to run the experiment
Copy NAME STATUS ROLES AGE VERSION
ip-mytesthost-w1.us-east-2.compute.internal Ready worker 4d5h v1.19.9
ip-mytesthost-w2.us-east-2.compute.internal Ready worker 4d5h v1.19.9
ip-mytesthost-m1.us-east-2.compute.internal Ready controlplane,etcd 4d5h v1.19.9
ip-mytesthost-m2.us-east-2.compute.internal Ready controlplane,etcd 4d5h v1.19.9
ip-mytesthost-w3.us-east-2.compute.internal Ready worker 4d5h v1.19.9
ip-mytesthost-m3.us-east-2.compute.internal Ready controlplane,etcd 4d5h v1.19.9 4. Initiate the redsky controller manager pod on the kubernetes cluster
Copy Warning: apiextensions.k8s.io/v1beta1 CustomResourceDefinition is deprecated in v1.16+, unavailable in v1.22+ ; use apiextensions.k8s.io/v1 CustomResourceDefinition
customresourcedefinition.apiextensions.k8s.io/experiments.redskyops.dev configured
customresourcedefinition.apiextensions.k8s.io/trials.redskyops.dev configured
clusterrole.rbac.authorization.k8s.io/redsky-manager-role configured
clusterrolebinding.rbac.authorization.k8s.io/redsky-manager-rolebinding unchanged
namespace/redsky-system unchanged
deployment.apps/redsky-controller-manager unchanged
clusterrole.rbac.authorization.k8s.io/redsky-patching-role unchanged
clusterrolebinding.rbac.authorization.k8s.io/redsky-patching-rolebinding unchanged
secret/redsky-manager configured 5. Verify if the redsky-controller-manager is running on k8s cluster
Copy kubectl get pod -n redsky-system
Copy NAME READY STATUS RESTARTS AGE
redsky-controller-manager-6cbb796b79-5wnz8 1/1 Running 1 4d4h 6. Authorize the kubernetes cluster, where user want to run the experiment and application is present
Copy redskyctl authorize-cluster
Copy secret/redsky-manager configured
deployment.apps/redsky-controller-manager patched Install Demo Application and Configure T4K Resources
1. There must be an application running in the namespace which will be used to perform the backup by StormForge experiment
Copy kubectl get pods | grep k8s-demo-app
Copy k8s-demo-app-frontend-7c4bdbf9b-bbz2z 1/1 Running 1 4d4h
k8s-demo-app-frontend-7c4bdbf9b-c84m7 1/1 Running 1 4d4h
k8s-demo-app-frontend-7c4bdbf9b-z8jhx 1/1 Running 1 4d4h
k8s-demo-app-mysql-754f46dbd7-v85mh 1/1 Running 2 4d4h 2. It is expected that Trilio for Kubernetes product is installed on the K8s cluster
Copy kubectl get pods | grep k8s-triliovault
Copy k8s-triliovault-admission-webhook-7c5b454ff4-85t92 1/1 Running 1 4d3h
k8s-triliovault-control-plane-54594f5796-vpz4r 2/2 Running 2 4d3h
k8s-triliovault-exporter-86d94c9967-bx56s 1/1 Running 1 4d3h
k8s-triliovault-ingress-gateway-6bdfc75c9b-l5qvb 1/1 Running 1 4d3h
k8s-triliovault-web-76d84fdcc5-lj52v 1/1 Running 1 4d3h
k8s-triliovault-web-backend-7fb8784588-8v865 1/1 Running 1 4d3h
triliovault-operator-k8s-triliovault-operator-66b6d9d96f-k8fjs 1/1 Running 1 4d4h 3. Verify that T4K License is applied to initiate the backup
Copy NAME STATUS MESSAGE CURRENT NODE COUNT GRACE PERIOD END TIME EDITION CAPACITY EXPIRATION TIME
test-license-1 Active Cluster License Activated successfully. 9 Basic 10 2022-04-30T00:0 4. Make sure that the Target to store the backup is configured and is in Available state
Copy NAME TYPE THRESHOLD CAPACITY VENDOR STATUS BROWSING ENABLED
demo-s3-target ObjectStore 100Gi AWS Available 5. Create a Backup plan for the application or namespace
Copy kubectl apply -f app/mysql-sample-backupplan.yaml
kubectl apply -f app/mysql-sample-backupplan-1.yaml 6. Verify the backup plan is created correctly
Copy kubectl get backupplan
Copy NAME TARGET RETENTION POLICY INCREMENTAL SCHEDULE FULL BACKUP SCHEDULE STATUS
mysql-label-backupplan demo-s3-target Available
mysql-label-backupplan-1 demo-s3-target Available Configure and Monitor StormForge Experiment
Users need to create a backup, bash script to monitor the backup and tvk-manager configmap.
Copy kubectl create configmap bashscript --from-file=configmaps/bash-script.sh
kubectl create configmap bashscript1 --from-file=configmaps/bash-script-1.sh
kubectl create configmap backup --from-file=configmaps/mysql-sample-backup.yaml
kubectl create configmap backup1 --from-file=configmaps/mysql-sample-backup-1.yaml
kubectl create configmap tvkmanager --from-file=configmaps/tvk-manager.yaml 2. User has all required entities in place to start the experiment
Copy kubectl apply -f experiment/tvk-scale-2-backup-delete-parallel-prod.yaml 3. User can monitor the running experiment being performed for each backup.
Copy kubectl get experiment
Copy NAME STATUS
tvk-scale-400-backup-delete-parallel-prod Running 4. Check the new Trial running by the experiment
Copy NAME STATUS ASSIGNMENTS VALUES
tvk-scale-400-backup-delete-parallel-prod-000 Setting up deploymentMemory=512, deploymentCpu=250, metaCpu=500, metaMemory=512, dataMemory=1536, dataCpu= 1200 5. Verify the jobs
Copy NAME COMPLETIONS DURATION AGE
tvk-scale-2-backup-delete-parallel-prod-001 0/1 4d3h 4d3h
tvk-scale-2-backup-delete-parallel-prod-001-create 1/1 51s 4d3h
tvk-scale-2-backup-delete-parallel-prod-001-delete 1/1 6s 4d3h Monitor the Experiment from StormForge UI
After the experiment is in running state, user can view the progress at the StormForge UI using the login they have configured.
2. Once the experiment run is complete, StormForge experiment will show the recommended configuration with resource details.
Delete / Remove StormForge Experiment
After the experiment run is complete, user can remove the StormForge redsky-controller-manager with simple commands
Delete the experiment from kubernetes cluster along with all other resources
Copy kubectl apply -f experiment/tvk-scale-2-backup-delete-parallel-prod.yaml Delete the experiment from redsky tenant
Copy redskyctl delete exp tvk-scale-2-backup-delete-parallel-prod Note: If you delete the experiment from redsky tenant, it will get deleted from the StormForge UI as well.
2. If user is facing any issues while running the experiment, they can check logs of redsky-controller-manager
Copy kubectl logs redsky-controller-manager-6cbb796b79-5wnz8 -n redsky-system Conclusion
Using StormForge, users can create desired experiments. Users run those experiments on Kubernetes cluster with applications running and perform operations like Backup/Restore. StormForge UI will provide the analysis through a well curated graph. It will show the best to worst combination of resources used to perform the operation in a particular trial. These optimized resource combinations can be applied in the production clusters to achieve the desired RPO/RTO.
