Fix: add design docs for ResourceTracker (#2909)

* Fix: enhance rt logic and add docs

Signed-off-by: Yin Da <yd219913@alibaba-inc.com>

* Fix: test conflict

Signed-off-by: Yin Da <yd219913@alibaba-inc.com>
This commit is contained in:
Somefive
2021-12-13 19:41:42 +08:00
committed by GitHub
parent c5c664f316
commit a89bb69a62
9 changed files with 184 additions and 12 deletions
@@ -0,0 +1,31 @@
# How to use ApplyOnce policy
By default, the KubeVela operator will prevent configuration drift for applied resources by reconciling them routinely. This is useful if you want to keep your application always have the desired configuration in avoid of some unintentional changes by external modifiers.
However, sometimes, you might want to use KubeVela Application to do the dispatch job and recycle job but want to leave resources mutable after workflow is finished. In this case, you can use the following ApplyOnce policy.
```shell
$ cat <<EOF | kubectl apply -f -
apiVersion: core.oam.dev/v1beta1
kind: Application
metadata:
name: apply-once-app
spec:
components:
- name: hello-world
type: webservice
properties:
image: crccheck/hello-world
traits:
- type: scaler
properties:
replicas: 1
policies:
- name: apply-once
type: apply-once
properties:
enable: true
EOF
```
In this case, if you change the replicas of the `hello-world` deployment after Application enters `running` state, it would be brought back. On the contrary, if you set the `apply-once` policy to be disabled (by default), any changes to the replicas of `hello-world` application will be brought back in the next reconcile loop.
@@ -1,4 +1,4 @@
## How to use garbage-collect policy
## How to keep legacy resources
Suppose you want to keep the resources created by the old version of the app. You only need to specify garbage-collect in the policy field of the app and enable the option `keepLegacyResource`.
@@ -0,0 +1,79 @@
# How to persist resources
By leveraging the garbage-collect policy, users can persist some resources, which skip the normal garbage-collect process when application is updated.
Take the following app as an example, in the garbage-collect policy, a rule is added which marks all the resources created by the `expose` trait to use the `onAppDelete` strategy. This will keep those services until application is deleted.
```shell
$ cat <<EOF | kubectl apply -f -
apiVersion: core.oam.dev/v1beta1
kind: Application
metadata:
name: garbage-collect-app
spec:
components:
- name: hello-world
type: webservice
properties:
image: crccheck/hello-world
traits:
- type: expose
properties:
port: [8000]
policies:
- name: garbage-collect
type: garbage-collect
properties:
rules:
- selector:
traitTypes:
- expose
strategy: onAppDelete
EOF
```
You can find deployment and service are created.
```shell
$ kubectl get deployment
NAME READY UP-TO-DATE AVAILABLE AGE
hello-world 1/1 1 1 74s
$ kubectl get service
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
hello-world ClusterIP 10.96.160.208 <none> 8000/TCP 78s
```
If you upgrade the application and use a different component, you will find the old versioned deployment is deleted by the service is kept.
```shell
$ cat <<EOF | kubectl apply -f -
apiVersion: core.oam.dev/v1beta1
kind: Application
metadata:
name: garbage-collect-app
spec:
components:
- name: hello-world-new
type: webservice
properties:
image: crccheck/hello-world
traits:
- type: expose
properties:
port: [8000]
policies:
- name: garbage-collect
type: garbage-collect
properties:
rules:
- selector:
traitTypes:
- expose
strategy: onAppDelete
EOF
$ kubectl get deployment
NAME READY UP-TO-DATE AVAILABLE AGE
hello-world-new 1/1 1 1 10s
$ kubectl get service
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
hello-world ClusterIP 10.96.160.208 <none> 8000/TCP 5m56s
hello-world-new ClusterIP 10.96.20.4 <none> 8000/TCP 13s
```