代码拉取完成，页面将自动刷新
The basic features have been completed, and while no breaking API changes are currently planned, the API can change in a backwards incompatible way before the project is declared stable.
Redis Operator can be considered a Kubernetes-native replacement for Redis Sentinel. It creates the Redis instances and maintains high availability and automatic failover.
Fundamental things to know about Redis Operator:
3 is a minimum number of Redis instances. Having 3 instances allows to always maintain a simple master-replica pair thus making it possible to replicate data even with Redis persistence turned off.role labels are added to each pod and end users are encouraged to use services to connect to master or replica nodes.Create a namespace for the operator:
kubectl create namespace redis-operator
Create all the necessary resources and deploy the operator:
kubectl apply -Rf deploy
Verify that the operator is running:
$ kubectl -n redis-operator get deployment
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
redis-operator 1 1 1 1 5m
Redis can be deployed by creating a Redis Custom Resource(CR).
Create a Redis CR that deploys a 3 node Redis replication in high availablilty mode:
kubectl create -f example/k8s_v1alpha1_redis_cr.yaml
Wait until the redis-example-... pods for the Redis replication are up and check the status of 'redis'. It will show the name for the Pod of the current master instance and the total number of replicas in the setup:
$ kubectl get redis example
NAME MASTER REPLICAS DESIRED AGE
example redis-example-0 3 3 24d
Scale the deployment:
$ kubectl scale redis example --replicas 4
redis.k8s.amaiz.com/example scaled
$ kubectl get redis example
NAME MASTER REPLICAS DESIRED AGE
example redis-example-0 4 4 24d
Redis Operator creates the following resources owned by the corresponding Redis CR. Please note that the name of Redis (example in this case) is used as an infix or suffix for the names of the generated resources:
redis-example (in case the password is set up)redis-exampleredis-exampleredis-exampleredis-example - covers all instancesredis-example-headless - covers all instances, headlessredis-example-master - service for access to the master instanceAll configuration of Redis is done via editing the Redis resourse file. Full annotated example can be found in the examples directory of the repo.
Delete the operators and CRDs. Kubernetes will garbage collect all operator-managed resources:
kubectl delete namespace redis-operator
kubectl delete crd redis.k8s.amaiz.com
The main goal of the project is to create a Kubernetes native service for managing highly available Redis deployments. Sentinel is great for simple Redis replication but it does not fit into Kubernetes model for a number of reasons:
Another imporatant goal of this project is to resist failures even with persistence turned off. In some scenarios persisting data on disk is not permitted and all the data should reside only in-memory no matter what. And at the same time losing this data is undesirable.
Redis Operator is not a distributed system. Instead it leverages the Kuberenetes API to perform a leader election upon startup. Current implementation of the leader election algorithm precludes the possibility of 2 instances mistakenly running as leaders (split brain).
Redis Operator watches for changes to the Redis resource as well as the resources managed by the operator and owned by Redis, e.g. ConfigMaps, Secrets, Services, PodDisruptionBudgets and StatefulSets. Should anything happen to any of the above the operator will check the state of the Kubernetes resources along with the state of Redis replication and reconfigure them if needed.
All the managed resources are created or updated in the first place. The resources already present are always compared to the resources generated by the operator and updated if they differ.
Once all the resources are in sync the list of Redis instances is compiled from the list of Pods owned by the corresponding StatefulSet. Only Pods with all containers running and ready are taken into account.
Minimum failover size is 2. 2 represents a simple master-replica pair essential for running replication. If the number of instances is less than the minimum failover size no reconfiguration will be performed. With this in mind it is absolutely normal to lose all instances but one at the same time. Even with persistence turned off the data will be preserved and replicated across all Pods that come in place of the terminated ones.
Redis Operator is stateless. It means that the state of replication is determined every time the list of Redis instances is analyzed.
The state of replication is determined based on the INFO REPLICATION output of every instance from the list.
A healthy replication is the state when there is a single master and all other instances are connected to it. In this case the operator will do nothing.
Master is an instance with at least one connected replica. If there is no masters found then there is one of two cases met:
In case the master has been lost the list of candidate instances are sorted according to their replica priority and replication offset. Instances with replica priority equal to 0 are filtered out prior to sorting.
A replica with the lowest priority and/or higher replication offset is promoted to master.
With the master in place all other instances that do not report themselves as the master's replicas are reconfigured appropriately. All replicas in question are reconfigured simultaneously.
Once the reconfiguration has been finished all Pods are labeled appropriately with role=master or role=replica labels. Current master's Pod name and the total quantity of connected instances are written to the status field of the Redis resource. The ConfigMap is updated with the master's IP address.
此处可能存在不合适展示的内容，页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容，可点击提交进行申诉，我们将尽快为您处理。
马建仓 AI 助手