Term
|
Definition
| kubectl get pods --selector app=app1 |
|
|
Term
|
Definition
- used for info only (not selectable)
- metadata:
labels:
app: app1
annotations:
version: v1 |
|
|
Term
|
Definition
- does NOT guarantee a specific node!
- if pod does NOT tolerate taint
- NoSchedule
- don't schedule pod ever
- PreferNoSchedule
- try not to schedule, but not guaranteed
- NoExecute
- do not schedule any new pods
- any pods already scheduled will be killed |
|
|
Term
|
Definition
- add
kubectl taint nodes "node" app=blue:NoSchedule
- remove
kubectl taint nodes "node" app=blue:NoSchedule- |
|
|
Term
|
Definition
spec:
tolerations:
- key:"app"
operator: "Equal"
value: "blue"
effect: "NoSchedule" |
|
|
Term
|
Definition
spec:
nodeSelector:
"label1": "value1" |
|
|
Term
|
Definition
| kubectl label nodes "node" key=value |
|
|
Term
| manually assign pod to node |
|
Definition
|
|
Term
|
Definition
spec:
affinity:
nodeAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
nodeSelectorTerms:
- matchExpressions:
- key: size
operator: In
values:
- Large
- Medium
OR
- matchExpressions:
- key: size
operator: NotIn
values:
- Small |
|
|
Term
|
Definition
- requiredDuringSchedulingIgnoredDuringExectuion
- no placement if no match
- if pod already assigned, ignore affinity
- preferredDuringSchedulingIgnoredDuringExecution
- try to match, but place anyway if no match
- if pod already assigned, ignore affinity |
|
|
Term
|
Definition
spec:
containers:
resources:
requests:
memory: 1Gi
cpu: 1
- cpu
0.1 = 100m = 100 mili
1 = 1 AWS vCPU
- memory
1G (1,000Kb) or 1Gi (1,024Kb)
or without suffix (just #'s) |
|
|
Term
|
Definition
- cpu = 1 or 1vCPU
- memory = 512Mi
spec:
containers:
resources:
limits:
memory:
cpu: |
|
|
Term
|
Definition
- cpu - throttle if pass limit
- memory - allow pass limit but kill container if continuously pass limit |
|
|
Term
|
Definition
apiVersion: v1
kind: LimitRange
metadata:
name: limit-range
spec:
limits:
- default:
memory: 512Mi
defaultRequest:
memroy: 256Mi
type: Container |
|
|
Term
| Running pods & deployments |
|
Definition
- Certain fields in pods cannot be changed when running so get yaml, change yaml, del pod, recreate pod
- Deployments can be edited when running and all pods will be deleted and recreated automatically |
|
|
Term
|
Definition
- Replicate pod to every node in cluster
- Replicate pod to new nodes in cluster
- Delete pods from deleted nodes in cluster
- Good for agents like monitoring and logging
- e.g. kube-proxy
- kubectl get daemonsets or ds |
|
|
Term
|
Definition
- same as replica set
- kind: DaemonSet |
|
|
Term
|
Definition
- kubelets read "pod-manifest-path" from kubelet.service file to get pod yaml files.
- used for master nodes to create control plane pods (apiserver, etcd, controller, etc.)
- pods can be viewed by apiserver, but read-only
- ignored by kube scheduler |
|
|
Term
|
Definition
spec:
schedulerName: "custom scheduler" |
|
|
Term
|
Definition
|
|
Term
|
Definition
| kubectl get pods --all-namespaces | grep "-master" |
|
|
