Let’s now create the Deployment using the kubectl apply command: $ kubectl apply -f deployment. k8s. StatefulSet. It specifies that there should be three replicas of the Deployment and that each pod should be labeled with app: nginx. To revert the deployment to the previous working version, use the rollout undo command: kubectl rollout undo statefulset/dbc1 -n mysql1 The output is similar to the following:Kubernetes Deployment vs StatefulSet: Which is Right for You? Kubernetes for Developers: Overview, Insights, and Tips. a cache or database cluster even though it may solve your specific situation as. metadata: name:. StatefulSets. This ensures it’ll be postgres-2 that’s destroyed first. But what is the best for this case ? This Pod is stateful (I am using volume hostPath to keep the data) and is deployed using nodeSelector to keep it always on the same Node. I'm using Logstash on Kubernetes and use the official Helm chart at this link. Manages the deployment and scaling of a set of Pods, and provides guarantees about the. The Reclaim Policy is used to determine the actions that need to be taken by the storage backend on deletion of the PV. Additionally, we are creating a Persistent Volume using the volumeClaimTemplate and using it in the StatefulSet to store the PostgreSQL data. Các Pod của Deployment cũng có thể bị xóa theo thứ tự bất kỳ, hoặc xóa đồng thời nhiều Pod (trong trường hợp scale down deployment) Statefulset. It gives the pods sequential names, starting with statefulsetname-0; and. Updated on: April 19, 2023 Sarav AK. Introduction A StorageClass provides a way for administrators to describe the "classes" of storage they offer. You should use a HorizontalPodAutoscaler (HPA for short): HPA automatically updates a workload resource (such as a Deployment or StatefulSet), with the aim of automatically scaling the workload to match demand. StatefulSet ensures that, at any time, there is at most one Pod with a given identity running in a cluster. Before you begin Before you begin this tutorial, you should familiarize yourself with the following Kubernetes concepts: Pods Cluster DNS Headless Services. The manifest files do not have the namespace added to them. 2. io/aws-ebs parameters: type: gp2 fsType: ext4 reclaimPolicy: Retain. To create the MongoDB replica set, run these two commands: kubectl apply -f googlecloud\_ssd. Logs stored persistently in storage, don't represent the state of the application. However, by defining a Deployment object, you can specify that Kubernetes should run multiple instances of the pod. Manages the deployment and scaling of a set of Pods, and provides guarantees about the ordering and uniqueness of these Pods. pods. yaml We need to create a Stateful yaml file for deploying MongoDB on Kubernetes. 04 Sep 2023 · 11 min read StatefulSets and Deployments are two Kubernetes API objects used to manage sets of identical Pods. At the highest level, a. Teams. Each StatefulSet domain will work as a DNS subdomain depending on the configuration of DNS for a Cluster. Deployment. Oct 12, 2020 Deployments vs StatefulSets in Kubernetes Deployments vs StatefulSets TL;DR Deployments are usually used for stateless applications while StatefulSets are. It makes sense also as each Redis instance relies on a configuration file that keeps track of other cluster instances and their roles. A StatefulSet is a controller for stateful applications, such as databases, and manages the deployment and scaling of pods. Define common group of shared volumes in Kubernetes (fsGroup) Define supplementalGroups inside Kubernetes SecurityContext. persistent buffering to disk), use a StatefulSet. g. kubectl create namespace database. io. To create a service, use the kubectl expose command. It also guarantees the uniqueness and ordering of the pods. A StatefulSet is a controller that helps you deploy and scale groups of Kubernetes pods. In my understanding JGroups Cluster is designed to work with stable node identities like they are provided by statefulset. When you have an app which requires persistence, you should create a stateful set instead of deployment. DaemonSet, Deployment, StatefulSet, and ReplicaSet: will no longer be served from extensions/v1beta1, apps/v1beta1, or apps/v1beta2 in v1. The Kubernetes StatefulSet controller gives each Pod a unique hostname based on its index. StatefulSet. We’re rapidly growing and always welcome new contributors. As this API is embedded in Pod's spec, you can use this feature in all the high-level workload APIs, such as Deployment, DaemonSet, StatefulSet, etc. In this guide, we explain to readers the differences between using a Kubernetes statefulset, versus using a deployment, as well as the use cases for each. Let’s rolling restart the ZooKeeper StatefulSet to update the pods to use the new version of the sidecar proxy: $ kubectl rollout restart statefulset my-release-zookeeper. Nevertheless, Kubernetes Operator does the same job. Unlike a Deployment, a StatefulSet maintains a. service没有ClusterIP，是headlessservice，所以无法负载均衡，返回的都是pod名，所以pod名字都. See full list on baeldung. ReplicationController is often abbreviated to "rc" in discussion, and as a shortcut in kubectl commands. Using a statefulset also ensures that. A Kubernetes Deployment YAML specifies the configuration for a Deployment object—this is a Kubernetes object that can create and update a set of identical pods. A Deployment works great for stateless applications where you can treat the pods as cattle. Kubernetes provides a basic resource called Pod. Advantages of using NFS with Kubernetes. I'm trying to define a shared persistent volume in k8s between two different deployments, and I've encountered some issues: I have 2 pods for each deployment and between the deployments I'm trying to configure a shared volume - that mean that if I create a txt file in deplyment1/pod1 and I take a look in deplyment1/pod2 - I can't see the file. It manages the deployment and scaling of a set of Pods, and provides guarantee about the ordering and uniqueness of these Pods. Finally, we will visualize and monitor all our data in. metadata. For example, you define how many replicas ( pods) of your app you want to run in the. The new pod will automatically spin up with correct configuration. In short, a pod is the core building block for running applications in a Kubernetes cluster; a deployment is a management tool used to control the way pods behave. 部署、扩展、更新、删除都要有顺序 2. Kubernetes administrators define classes of storage, and then pods can dynamically request the specific type of storage they need. Step-6: Updating the Secret with a new root password. 25. Unlike a Deployment, a StatefulSet. Deployment is a good fit for managing a stateless application workload on your cluster, where any Pod in the Deployment is interchangeable and can be replaced if needed. service. 2. Read more: Kubernetes StatefulSet: A Practical Guide. This page shows how to run a replicated stateful application using a StatefulSet. Like a Deployment, a StatefulSet manages Pods that are based on an identical container. They use a K8S component called Service, to group the pods using labels. Parallel. The problem is : a Pod will not be re-created after the Node failure (like a reboot). Of course, the scaling depends on the app you are deploying. This page provides an overview of init containers: specialized containers that run before app containers in a Pod. local however, these names are useless for me because I cannot set them in advance on my YAML files because it depends on the IP. PersistentVolumes and StatefulSets are the main approaches for running stateful applications in Kubernetes. Version the ConfigMap. I was playing with k8s deployment - rolling update and it works really well. This is referred to as at most. io/v1beta1 and CertificateSigningRequest to two keys. The last section of the manifest defines our StatefulSet. StatefulSetは状態を保持する (ステートフルな)アプリケーションを管理するためのKubernetesオブジェクトだ。. Gère le déploiement et la mise à l'échelle d'un ensemble de Pods, et fournit des garanties sur l'ordre et l'unicité de ces Pods. That's it! With these two commands, you have launched all the components required to run an highly available and redundant MongoDB replica set. 10. The team needs to compose a new image for each project to avoid confusion in executing commands. Contrairement à un Deployment, un. Four Pods are running. If, in any case, one of the pod die,s the newly generated pod will be of the same ID. It will trigger them all at once. Deployments allow you to manage sets of identical pods (or ReplicaSets) using common configurations. StatefulSet is the workload API object used to manage stateful applications. The execution of a Pod depends on other ordinal index. 2. This is a living document. Manages the deployment and scaling of a set of Pods, and provides. Related Resources. StatefulSet is a Kubernetes controller used to deploy stateful apps. conditions[1): unknown field "lastUpdateTime" in io. Specifying minReadySeconds slows down a rollout of a StatefulSet, when using a RollingUpdate value in . Kubectl is the command line configuration tool for Kubernetes that communicates with a Kubernetes API server. NetApp Trident is a dynamic container storage interface (CSI) for use in Kubernetes. replicas. 22 introduced a new ReadWriteOncePod access mode for PersistentVolumes and PersistentVolumeClaims. io. To check the image deployed on these Pods, use the describe pods subcommand: kubectl describe pods. En esta entrada vamos a tratar sobre aplicaciones Stateful en Kubernetes – Statefulset, el cual es un tipo de recurso que nos ofrece Kubernetes. Full k8s cluster consists of 4–5 services on the control plane and two on worker nodes. SQL Server is a database application and thus mostly should be deployed as a StatefulSet workload type. Unlike a Deployment, a StatefulSet maintains a sticky identity for each of their Pods. Manages the deployment and scaling of a set of Pods, and provides guarantees about the ordering and uniqueness of these Pods. quiz across different areas of your DevOps process (workflows, deployments, team collaboration) to see how you score vs. In a stateful containerized application, data must be persistent, retained and easy to access outside the application. 1. Quick tutorial #2: Creating an NFS Persistent Volume. For general information about working with config files, see deploying applications, configuring containers, managing resources. It's created after deployment. Latest Deployment Generation: kube_deployment_metadata_generation: Sequence number representing the latest generation of a Deployment. updateStrategy. Labels can be used to select objects and to find collections of objects that satisfy certain conditions. When a pod instance managed by a StatefulSet disappears (because the node the pod was running on has failed, it was evicted from the node, or someone deleted the pod object manually), the StatefulSet makes sure it’s. Stateful app pods require unique identities because they communicate with other pods. Moreover, if at any point one of the replicas fails to start, the whole process will stop. k8s. Product. service没有ClusterIP，是headlessservice，所以无法负载均衡，返回的都是pod名，所以pod名字都. g. Apply and Delete Kubernetes resource files. StatefulSet. StatefulSetCondition] You should delete everything from the status field. This is important because many stateful applications have data initialization routines to perform, masters to elect, and quorum to achieve. This controller monitors the state of Kubernetes Deployments, the most common approach to deploying a workload in Kubernetes. pod名字始终是固定的 4. Yes, my metrics server is running fine. This means that each pod has its own IP address, making. So, for example, Kubernetes will only create example-statefulset-5 after example-statefulset-4 is up and running. Al igual que un Deployment, un StatefulSet gestiona Pods que se. It is mainly designed to use for stateful apps. Ordinals can start from arbitrary non-negative numbers. Deployments Deployment is the easiest and most used resource for deploying an application. A Kubernetes pod is a cluster deployment unit that typically contains one or more containers. Each pod runs specific containers, which are defined in the spec. Quick tutorial #1: Mounting an NFS share on a container. If you think of something that is not on this list but might be useful to others, please don't hesitate to file an issue or submit a PR. A StatefulSet is a workload object that makes it possible to deploy Pods (which host Kubernetes workloads) with unique identifiers – which you don't get from a standard Kubernetes deployment. api. kubectl basics. Learn more about TeamsAdding entries to a Pod's /etc/hosts file provides Pod-level override of hostname resolution when DNS and other options are not applicable. Above is one example of a Deployment Strategy called Blue/Green. Deployment vs StatefulSet component Pods deployed by Deployment are identical and interchangeable, created in random order with random hashes in their Pod. En esta entrada vamos a tratar sobre aplicaciones Stateful en Kubernetes – Statefulset, el cual es un tipo de recurso que nos ofrece Kubernetes. yaml) and paste in the following configuration settings:nodeSelector can not be used in persistentVolumes. The hostnames take the form of <statefulset name>-<ordinal index>. As best we can tell, Nextcloud unfortunately is a stateful application. storage. 1 Answer. Create a StatefulSet. Kubernetes Apply. This is important because many stateful applications have data initialization routines to perform, masters to elect, and quorum to achieve. template field of the YAML configuration. To check your version of Kubernetes, run kubectl version. First, you can list the pods to identify the ones you want to restart: kubectl get podsIt worked as expected and both the prometheus pods got restarted. This comes at a cost of slow scale ups and. The last step is to "assign" volume to your execution (Pod, Deployment, StatefulSet, etc) which is done using volumes. 2. Also, you will not have to create a PVCs in advance, and you will be able to scale it easily. StatefulSets. So, to create a dummy deployment we use: bash. Resource Objects. g. As the name suggest the statefulsets are specifically used to make sure the app state is saved by each running pod. nodeSelector should be in the spec section of pod template. [root@controller ~]# kubectl create deployment nginx-deploy --image=nginx --dry-run=client -o yaml > nginx-deploy. Use-cases: Daemonset vs StatefulSet. Waypoint helps simplify the deployment process with its Helm plugin and does so automatically with its Git integration and GitOps. Pods created as part of a StatefulSet are given persistent. When you use a StatefulSet, Kubernetes terminates Pods in the opposite order to their creation. moury~ % kubectl scale statefulset web --replicas=3 statefulset. kubectl api-resources -o wide | grep -i deployment will provide the relevant information. 16. DeploymentはRollingUpdateとParallelの2つのストラテジーがあります。You can also update the object's resource requests and limits, labels, and annotations. The application is MySQL. If this is a serious project you should definitely change it to a statefulset. These pods are created from the same specification, but they are not interchangeable: each one has a persistent identifier maintained across any. Pods (and, by extension, containers) are, nevertheless, short-lived entities. Name Stays the Same. Statefulset es un recurso que controla y gestiona en kubernetes uno o más Pods, pero entonces por qué no usar un Deployment te estarás preguntando?, entonces antes de continuar es mejor que veas las diferencias entre Stateful y Stateless en una. FEATURE STATE: Kubernetes v1. It is responsible for the deploying, scaling of a set of Pods, and guarantees the ordering and uniqueness of these Pods. If you look at web_stateful. A DaemonSet is a unique kind of resource that K8s assigns to a pod for each Kubernetes node in the cluster. Pods created as part of a StatefulSet are given persistent identifiers that they retain even when they’re rescheduled. Các Pod của Statefuleset không hoàn toàn giống. Các Pod của Statefuleset không hoàn toàn giống. spec. The first line of this snippet copies the original StatefulSet and creates a statefulSetCanary, which has some additional adjustments, like the `-canary` suffix, the `track: canary` label, and. Example-2: Define runAsUser for container. Deployments and ReplicationControllers are meant for stateless usage and are rather lightweight. First, do a git pull to ensure you’re working on the latest revision: $ git pull origin setup-semaphore. Community. Deployment vs StatefulSet component. $ kubectl -n=mehdb apply -f app. This ensures it’ll be postgres-2 that’s destroyed first. yml. StatefulSet Deployments provide: Stable, unique network identifiers: Each pod in a StatefulSet is given a hostname that is based on the application name and increment. Here is the quote from a relevant section from the docs: Like a Deployment, a StatefulSet manages Pods that are based on an identical container spec. For example, an application that writes logs to Persistent Volume claim (PVC) in K8s could be considered as stateless even using storage. Whilst a Pod is running, the. When pods run as part of a StatefulSet, Kubernetes keeps state data in the persistent storage volumes of the StatefulSet, even if the pods shut down. 1. Run a Stateless Application Using a Deployment; Run a Single-Instance Stateful Application; Run a Replicated Stateful Application; Scale a StatefulSet; Delete a. The deployment is made up of 2 Nginx web servers; both of them are connected to a persistent volume. Horizontal scaling means that the response to increased load is to deploy more Pods. Therefore the latter use. If it replaces them because the configuration changes, it keeps the names the same. Unlike a. Our Mongodb Setup will be as follows : 1x Config Server (k8s deployment type: “StatefulSet”)Stateful Applications in Kubernetes. 目前看来有点本末倒置，不过不. Caching Systems: StatefulSets can be used to deploy caching systems in Kubernetes, as they provide stable network identities, persistent storage, and predictable deployment and scaling. Note: Pods can only reference imagePullSecrets in their own namespace, so this. k8s. Deployment manages creating Pods by means of ReplicaSets. Updates are versioned and you can revert to any previously known state of a Deployment. To run a Redis cluster on Kubernetes we will use a StatefulSet. Cloud Observability recommends running the OpenTelemetry Collector with the Prometheus receiver to ingest infrastructure metrics. Each individual deployment is represented as a replication controller. You should read about Kubernetes autoscaling - HPA. (K8S default), set the image of the container to a new version for a particular deployment. The rollout process is a gradual, step-by-step recreation of Pods, or instances of an application, one by one. Create a service to expose the PostgreSQL database within the Kubernetes cluster, then create a file (e. For example, you can use the DaemonSet to scrape application metrics along with a single. 7 Answers. For simplicity, are we using Nginx as the pod image. Deployment vs. 每个pod都有自己存储，所以都用volumeClaimTemplates，为每个pod都生成一个自己的存储，保存自己的状态 3. network issue), for Deployment, a new Pod will be created on a different node (to follow your desired 1 replica), but for StatefulSet it will make sure to terminate the existing Pod before creating a new, so that there are never more than 1 (when you have 1 as desired number of replicas). - Storage: As many VolumeClaims as requested. Both Deployment and StatefulSet try to maximize the availability - but StatefulSet cannot sacrifice data consistency (e. apps. Deployments vs Daemonsets vs Statefulsets. A pod is the smallest deployable unit in Kubernetes which has a collection of containers. pod名字始终是固定的 4. k8s. In a stateful containerized application, data must be persistent, retained and easy to access outside the application. your state), so it cannot act as fast as Deployment (stateless) apps can. StatefulSet. Kubernetes users don’t need to worry about how pods are scheduled—they can deploy pods sequentially, attach them to persistent storage volumes, and each pod maintains its own persistent network ID. A StatefulSets are Kubernetes objects used to consistently deploy stateful application components. In Kubernetes, a HorizontalPodAutoscaler automatically updates a workload resource (such as a Deployment or StatefulSet), with the aim of automatically scaling the workload to match demand. Por ejemplo, si creamos un StatefulSet. StatefulSets also enforce that Pods are removed in reverse order of their creation. When using Kubernetes, most of the time you don’t care how your pods are scheduled, but sometimes you care that pods are deployed in order, that they have a persistent storage volume, or that they have a unique, stable network identifier across. StatefulSetの概要. A Pod's contents are always co-located and co-scheduled, and run in a. Otherwise you can use Deployments with multiple pods online for your shards. This is great for stateless apps that we want online, ASAP. A Kubernetes StorageClass is a Kubernetes storage mechanism that lets you dynamically provision persistent volumes (PV) in a Kubernetes cluster. Recreate Strategy. Even when there is a new rolling update, the statefulset still in the same state CrashLoopBackOff from the old rolling update. Comme un Déploiement, un StatefulSet gère des Pods qui sont basés sur une même spécification de conteneur. Familiarity with volumes and persistent volumes is suggested. Step 3: Create and deploy the StatefulSet. Deployment modes A bare metal deployment of the collector is simple to plan and execute: it’s a single binary that runs as a daemon on the host. Let’s take a look at Deployments, DaemonSets, and StatefulSets. StatefulSetの概要. At this point, the Green Pods retire, and Blue becomes the new Green. In contrast, annotations are not used to identify and select objects. A simple case is to create one ReplicationController object to reliably run one instance of a Pod indefinitely. Pods are the smallest deployable units of computing that you can create and manage in Kubernetes. It demonstrates how to create, delete, scale, and update. kubernetes. A service is responsible for enabling network access to a set of pods. You can use either labels or annotations to attach metadata to Kubernetes objects. This comes at a cost of slow scale ups and. Before you begin This is a fairly advanced task and has the potential to violate some of the properties inherent to StatefulSet. a cache or database cluster even though it may solve your specific situation as. @Jarek the question was more whether you can technically use a normal service for a StatefulSet. Author: Chris Henzie (Google) Last month's release of Kubernetes v1. Attributes of a Highly Available WordPress Deployment. Apache Spark is a stateful service, those should be deployed as StatefulSet. This is where PersistentVolumes come into play. Like. I also show you how to deploy databases using. 0}. DaemonSets. This is where PersistentVolumes come into. For example, you define how many replicas ( pods) of your app you want to run in the deployment. fluentd-elasticsearch namespace: kube-system labels: k8s-app: fluentd-logging spec: selector: matchLabels: name. 2. Kubernetes deployments manage stateless services running on your cluster (as opposed to for example StatefulSets which manage stateful services). Priority indicates the importance of a Pod relative to other Pods. 15. g. One pod should be able to reach other pods with well-defined names. updateStrategy field, by waiting for each pod for a. Read more about StatefulSet here. “Kubernetes — Difference between Deployment and StatefulSet in K8s” is published by Ashish Patel. yaml. A StatefulSet is a workload API object for managing stateful applications. requires writes. This controller runs Jobs -- components of a Kubernetes workload that execute specific tasks -- according to. Kubernetes is a powerful container orchestration tool that provides a variety of resources for managing containerized applications. You can add these custom entries with the HostAliases field in PodSpec. Storage classes can define properties of storage systems. Kubernetes deployments manage stateless services running on your cluster (as opposed to for example StatefulSets which manage stateful services). Then, the second pod (2) does likewise. This document highlights and consolidates configuration best practices that are introduced throughout the user guide, Getting Started documentation, and examples. Rolling Update Strategy. The . Introduction to autoscaling in k8s. It is resilient against node failures and ensures appropriate data archiving. vim redis-statefulset. your peers. There is an entry for resources in my statefulset yaml and CPU limit is set to 1 and request to 0. StatefulSets are designed to run stateful applications in Kubernetes with dedicated persistent storage. This page shows how to delete Pods which are part of a stateful set, and explains the considerations to keep in mind when doing so. We assign it a service name, specify a single replica,. When using Kubernetes, most of the time you don’t care how your pods are scheduled, but sometimes you care that pods are deployed in order, that they have a persistent storage volume, or that they have a unique, stable network identifier across restarts. v1. The StatefulSet controller provides each Pod with a unique hostname based on its ordinal index. In statefulsets each replica pod created has an index number starting from 0 and it will only setup the next replica if the previous one is running. StatefulSet is the workload API object used to manage stateful applications. The solution(s) : Use a StatefulSet, ReplicaSet or DaemonSet to ensure the Pod creation after a Node failure. 1. 27, this feature is now beta. Pods (and, by extension, containers) are, nevertheless, short-lived entities. If a Pod cannot be scheduled, the scheduler tries to preempt (evict) lower priority Pods to make scheduling of the pending Pod possible. The common format of a kubectl command is: kubectl action resource This performs the specified action (like create, describe or delete) on the specified resource (like node or deployment). If there's heavy load, you can set up a HorizontalPodAutoscaler to. yaml You should receive the following output: service/mongo created statefulset. First, find the StatefulSet you want to scale. First sentence in the documentation: "The Horizontal Pod Autoscaler automatically scales the number of Pods in a replication controller, deployment, replica set or stateful set". Deployment vs StatefulSet. 26 introduced a new, alpha-level feature for StatefulSets that controls the ordinal numbering of Pod replicas. A Replica set is used to ensure that a specific set of Pods is running at all times. Unlike a Deployment, a StatefulSet has a persistent ID for each Pod that persists across any rescheduling or. Deploying a Stateful Application Using Kubernetes Statefulset. selector. The above command will create a ReplicaSet with three replicas and manage the lifecycle of the pods. . Monitoring is similar, though — you need to make sure that the number of desired StatefulSet Replicas matches the number of ready StatefulSet Replicas. Deployment vs StatefulSet. You need to mount the configMap rather than use it as an environment variable, as the setting is not a key-value format. apps. In this example: A Deployment named nginx-deployment is created, indicated by the . First, we will create a persistent volume claim for the storage we need. Read more about StatefulSet here. A StatefulSet is another Kubernetes controller that manages pods just like Deployments. Un StatefulSet es el objeto de la API workload que se usa para gestionar aplicaciones con estado. Your Deployment yaml should be like below: containers: - name: nginx image: nginx volumeMounts: - mountPath: /etc/nginx name: nginx-conf volumes: - name: nginx-conf configMap: name: nginx-conf items: - key:. If you don’t specify the namespace, it gets deployed in the default namespace. We usually need StatefulSet when we have a stateful app, so every pod can have its own volume. How to Use kubectl rollout restart. The most basic difference is that you would get ability to persist pod level state with statefulsets. Deployment. Deleting and/or scaling a StatefulSet down will not delete the volumes associated with the. We’re rapidly growing and always welcome new contributors. All the examples available are showing Redis cluster being deployed as a combination of Kubernetes’ StatefulSets and PersistentVolumes. Unlike a. . This page describes the lifecycle of a Pod. CPU requests are mostly more important for the kube-scheduler to identify the best node suitable to place a pod. See StatefulSet vs. yaml, we can run the following command to run both of them: kubectl apply -f mongo-statefulset. deepak. statefulsets do ordinal scheduling; first pod 0, then pod 1, etc. Then, the second pod (2) does likewise. If you don’t specify the namespace, it gets deployed in the default namespace. Then we run the command for the deployment template : kubectl apply -f demo-deployment. Securely running workloads in Kubernetes can be difficult. In addition to kubectl rollout restart deployment, there are some alternative approaches to do this: 1. Their purpose is to keep a set of identical pods running and upgrade them in a controlled way.