Replicationcontroller and replicaset of kubernetes

kubernetes And ReplicationController And ReplicaSet

ReplicationController Part of the

One ReplicationController There are three main parts ：

• label selector（ tag chooser ）： Used to determine the ReplicationController What's in scope pod.
• replica count（ Number of copies ）： Specify the pod Number .
• pod template（pod Templates ）： Used to create a new pod copy .

The above three parts can be modified at any time , among replica count That is, the change of the number of copies will affect the existing pod. Change the label selector
Will make the existing pod Out of the ReplicationController The scope of the , So the controller stops paying attention to them .

Use ReplicationController The benefits of

• Ensure a pod Or more pod The copy runs continuously , The way is now pod Start a new one when lost pod.
• When a cluster node fails , It will be all of the... Running on the failed node pod（ Immediate acceptance ReplicationController On the control node pod） Create alternate copies .
• It can be easily realized pod The horizontal expansion of .

edit ReplicationController

View owned rc：kubectl get rc

edit rc：kubectl edit re kubia

Delete ReplicationController And keep pod The normal operation

When using kubectl delete Delete ReplicationController when , You can add --cascade=false Options to keep pod function .
namely ：kubectl delete rc kubia --cascade=false

Use ReplicaSet instead of ReplicationController

ReplicaSet It's a new generation of ReplicationController, And completely replace the old , namely ReplicationController Will eventually be discarded .
That is to say, we should create ReplicaSet instead of ReplicationController, They're almost identical , So there will be no learning cost .

ReplicaSet What are the advantages of ？
Its pod Selectors are more expressive . for instance , Single ReplicationController Cannot be pod And labels env=a And env=b Simultaneous matching , It can only match with
Yes env=a Label or env=b Labeled pod. but ReplicaSet Can match two groups pod And treat them as a large group .
Can pass matchExpressions Property to override the selector , Such as some examples ：

selector:
  matchExpressions:
    - key: app #  This selector requires this pod Contains the name app The label of 
      operator: In
      values:
        - dev #  The tag value must be values value ： namely dev

Example above , Each expression must contain a key, One operator（ Operator ）, And there may be another values A list of （ Depends on the running symbol ）.
Valid running symbol ：

• In：Label The value of must match one of the specified values matching .
• NotIn：Label With any specified value values Mismatch .
• Exists：pod Must contain a label with the specified name , Value doesn't matter , When using this operator , You should not specify values Field .
• DoesNotExist：pod Must not contain a label with the specified name .values Property must not specify .

It should be noted that , If more than one expression is specified , Then all of these expressions must be true To make the selector and pod matching . If you also specify matchLabels and matchExpressions,
Then all tags must match , And all expressions must be calculated as true, To make this pod Match selector .

How to define ReplicaSet？
The same is to create yaml file , But here's the thing ReplicaSet No V1 API Part of , But it belongs to apps API Of the group v1beta2 edition . So you need to make sure that in
Specify the correct... When creating resources apiVersion. except kind Need by ReplicationController Change to ReplicaSet, Other content and creation ReplicationController Almost the same .

It is worth noting that ReplicaSet The abbreviation is rs. May be through kubectl get rs Check it out. ReplicaSet.

Reference books ：《Kubernetes in action》