Resource Watcher

Introduction

An incident response if delayed can impact businesses, revenue, and waste valuable engineering time. Devtron's Resource Watcher enables you to perform automated actions upon the occurrence of events:

• Create Event - Occurs when a new Kubernetes resource is created, for e.g., a new pod spun up to handle increased traffic.

• Update Event - Occurs when an existing Kubernetes resource is modified, for e.g., deployment configuration tweaked to increase the replica count.

• Delete Event - Occurs when an existing Kubernetes resource is deleted, for e.g., deletion of an orphaned pod.

You can make the Resource Watcher listen to the above events and accordingly run a job you wish to get done, for e.g., increasing memory, executing a script, raising Jira ticket, emailing your stakeholders, sending Slack notifications, and many more. Since manual intervention is absent, the timely response of this auto-remediation system improves your operational efficiency.

Creating a Watcher

This page allows you to create a watcher to track events and run a job. It also shows the existing list of watchers (if any).

1. Click + Create Watcher.

   
2. Creating a watcher consists of 4 parts, fill all the sections one by one:

   • Basic Details

   • Namespaces to Watch

   • Intercept Change in Resources

   • Execute Runbook

   

Basic Details

Here, you can give a name and description to your watcher.

Namespaces to Watch

Here, you can select the namespaces whose Kubernetes resource you wish to monitor for changes.

• You can watch the namespace(s) across All Clusters (existing and future).

  
• Or you can watch namespace(s) of Specific Clusters.

  

Intercept Change in Resources

Here, you can select the exact Kubernetes resource(s) you wish to track for changes (in the namespace(s) you selected in the previous step).

• You can choose the resource from the Resource kind(s) to watch dropdown. Enter the Group/Version/Kind (GVK) if it's a custom resource definition (CRD), for e.g., install.istio.io/v1apha1/IstioOperator

• Choose the event type your watcher should listen to: Created, Updated, Deleted.

  Event Type

  Description

  Created

  Triggers the watcher when your Kubernetes resource is created

  Updated

  Triggers the watcher when your existing Kubernetes resource is modified

  Deleted

  Triggers the watcher when your existing Kubernetes resource is deleted

• Enter a CEL expression to catch a specific change in the resource's manifest.

Example: DEVTRON_FINAL_MANIFEST.status.currentReplicas == DEVTRON_FINAL_MANIFEST.spec.maxReplicas

Execute Runbook

Here, you can choose a job that should trigger if your watcher intercepts any changes.

• Choose a job pipeline from the Run Devtron Job pipeline dropdown. If a pipeline is not selected, the watcher won't intercept matching resource changes even if your defined conditions are met.

• Select the environment in which the job should run. It can either be devtron-ci or the source environment (the intercepted namespace where the event has occurred).

• If the job expects input parameters, you may add its key and value under Runtime input parameters.

  During a job's execution, its container can access the initial and final resource manifest through special environment variables. These variables are:

  • DEVTRON_INITIAL_MANIFEST

  • DEVTRON_FINAL_MANIFEST

• Click Create Watcher.

Your watcher is now ready to intercept the changes to the selected resources.

Viewing Intercepted Changes

Details

This page allows you to view the changes to Kubernetes resources that you have selected for tracking changes.

It comes with the following items to help you locate the resource, where the event has been intercepted:

• Searchbox

• Cluster filter

• Namespace filter

• Action filter (event type, i.e., Created, Updated, Deleted)

• Watcher filter (to check the intercepted changes of a specific watcher)

You get the following details in the results shown on the page.

Change in Resource

You can check the changes in manifest by clicking View Manifest in Change In Resource column.

Job Execution Log

You can check the logs of the job executed when the Resource Watcher intercepts any change by clicking logs.

Use Cases

Live Stream Traffic Surge

A live streaming sports application experiences a surge in viewers during a major game. The Horizontal Pod Autoscaler (HPA) might not be able to handle the unexpected traffic if it's capped at a low max replica count.

1. Create a watcher named 'Live Stream Scaling Alert'.

2. Monitor updates to HPA resource in the application's namespace.

3. When currentReplicas count reaches maxReplicas, trigger a job that contains the script to increase the replica count.

Pod Health Monitoring

A stock trading application constantly updates stock prices for its traders. If the pods become unhealthy, traders might see incorrect stock prices leading to bad investments.

1. Create a watcher named 'Pod Health Monitor'.

2. Track the pod workload of your application, if DEVTRON_FINAL_MANIFEST.status.phase != 'Running', trigger a job that sends an Email/Slack alert with pod details.