- Git
- Golang (version >= 1.13)
- Docker (version >= 19.03)
- Kubernetes (version >= 1.14)
- GNU Make
For installation of Golang, please refer to Install Golang
make
is usually in a build-essential
package in your distribution's package manager of choice. Make sure you have make
on your machine.
There're great chances that you may want to run your implementation in a real Kubernetes cluster, so probably a Docker is needed for some necessary operations like building images. See Install Docker for more information.
$ mkdir -p $GOPATH/src/github.com/fluid-cloudnative/
$ cd $GOPATH/src/github.com/fluid-cloudnative
$ git clone https://github.com/fluid-cloudnative/fluid.git
NOTE: In this document, we build, run and debug under non-module environment.
See Go Modules for more information if some issue occurs to you.
Makefile
under project directory provides many tasks you may want to use including Test, Build, Debug, Deploy etc.
You can simply get a binary by running:
# build dataset-controller, alluxioruntime-controller and csi Binary
$ make build
# if you only want to build alluxioruntime-controller
$ make alluxioruntime-controller-build
By default, the binary would be put under <fluid-path>/bin
.
-
Set tags for images
# set name for image of dataset-controller $ export DATASET_CONTROLLER_IMG=<your-registry>/<your-namespace>/<img-name> # set name for image of alluxioruntime-controller $ export ALLUXIORUNTIME_CONTROLLER_IMG=<your-registry>/<your-namespace>/<img-name> # set name for image of CSI $ export CSI_IMG=<your-registry>/<your-namespace>/<csi-img-name> # set name for image of init-user $ export INIT_USERS_IMG=<your-registry>/<your-namespace>/<csi-img-name> # build all images $ make docker-build-all # build and push all images using docker buildx (Run QEMU emulation before docker buildx, as below) $ docker run --rm --privileged multiarch/qemu-user-static --reset -p yes $ make docker-buildx-all-push
Before running Fluid, you need to push the built image to an accessible image registry.
-
Login to a image registry
Make sure you've login to a docker image registry that you'd like to push your image to:
$ sudo docker login <docker-registry>
-
push your images:
$ make docker-push-all
In the following steps, we assume you have properly configured KUBECONFIG
environment variable or set up ~/.kube/config
. See Kubeconfig docs for more information.
-
Push your images to a image registry accessible to your Kubernetes cluster
If your images are pushed to some private repositories, make sure your Kubernetes cluster hold credentials for accessing those repositories.
-
Change image in the samples we provide:
# <fluid-path>/config/fluid/patches/image_in_manager.yaml ... ... containers: - name: manager image: <registry>/<namespace>/<img-repo>:<img-tag>
# <fluid-path>/config/fluid/patches/image_in_csi-plugin.yaml ... ... containers: - name: plugins image: <registry>/<namespace>/<csi-img-name>:<csi-img-tag>
-
Install CRDs
$ kubectl apply -k config/crd
Check CRD with:
$ kubectl get crd | grep fluid alluxiodataloads.data.fluid.io 2020-08-22T03:53:46Z alluxioruntimes.data.fluid.io 2020-08-22T03:53:46Z datasets.data.fluid.io 2020-08-22T03:53:46Z
-
Install your implementation
$ kubectl apply -k config/fluid
Check Fluid system with:
$ kubectl get pod -n fluid-system NAME READY STATUS RESTARTS AGE controller-manager-7fd6457ccf-p7j2x 1/1 Running 0 84s csi-nodeplugin-fluid-pj9tv 2/2 Running 0 84s csi-nodeplugin-fluid-t8ctj 2/2 Running 0 84s
-
Run samples to verify your implementation
Here is a sample provided by us, you may want to rewrite it according to your implementation.
$ kubectl apply -k config/samples
Check sample pods:
$ kubectl get pod NAME READY STATUS RESTARTS AGE cifar10-fuse-vb6l4 1/1 Running 0 6m15s cifar10-fuse-vtqpx 1/1 Running 0 6m15s cifar10-master-0 2/2 Running 0 8m24s cifar10-worker-729xz 2/2 Running 0 6m15s cifar10-worker-d6kmd 2/2 Running 0 6m15s nginx-0 1/1 Running 0 8m30s nginx-1 1/1 Running 0 8m30s
-
Check logs to verify your implementation
$ kubectl logs -n fluid-system <controller_manager_name>
-
Clean up
$ kubectl delete -k config/samples $ kubectl delete -k config/fluid $ kubectl delete -k config/crd
Execute following command from project root to run basic unit tests:
$ make unit-test
kubebuilder
provided a integration test framework based on envtest package. You must install kubebuilder
before running integration tests:
$ os=$(go env GOOS)
$ arch=$(go env GOARCH)
$ curl -L https://go.kubebuilder.io/dl/2.3.1/${os}/${arch} | tar -xz -C /tmp/
$ sudo mv /tmp/kubebuilder_2.3.1_${os}_${arch} /usr/local/kubebuilder
$ export PATH=$PATH:/usr/local/kubebuilder/bin
Next, run all unit tests (integration tests included) with:
$ make test
NOTE: When running unit tests on a non-linux system such as macOS, if testing failed and says
exec format error
, you may need to check whetherGOOS
is consistent with your actual OS.
The Fluid controller component supports local operation or debugging. The Fluid controller components include Dataset Controller, various runtime controllers, and Application Controller. Before running the controller component locally, it is necessary to configure kubeconfig in advance in the local environment (configured through the KUBECONFIG
environment variable or through the $HOME/.kube/config
file) and be able to access a Kubernetes cluster normally.
The Fluid CSI plugin cannot run locally to interact with the Kubernetes cluster. To run such components, it is necessary to first perform image construction, manually replace the corresponding image address of
charts/fluid/fluid/values.yaml
, and then deploy it to the Kubernetes cluster.
The runtime controller depends on heml
. And related Helm Charts to function properly. Therefore, before running the Runtime Controller, execute the following command to configure the environment:
- Create a soft link for the helm command locally
$ ln -s $(which helm) /usr/local/bin/ddc-helm
- Create a soft link directory for related Charts locally
$ ln -s $GOPATH/src/github.com/fluid-cloudnative/fluid/charts $HOME/charts
- Taking the Alluxio Runtime Controller as an example, run the component locally using the following command:
# Configure environment variable parameters related to AlluxioRuntime
$ export ALLUXIO_RUNTIME_IMAGE_ENV="alluxio/alluxio-dev:2.9.0"
$ export ALLUXIO_FUSE_IMAGE_ENV="alluxio/alluxio-dev:2.9.0
$ export DEFAULT_INIT_IMAGE_ENV="fluidcloudnative/init-users:v0.8.0-5bb4677"
$ export MOUNT_ROOT="/runtime-mnt"
$ export HOME="$HOME"
# Open the development debugging mode, open leader-election, and start alluxioruntime-controller
$ ./bin/alluxioruntime-controller start --development=true --enable-leader-election
The Fluid controller component supports local operation or debugging. The Fluid controller components include Dataset Controller, various runtime controllers, and Application Controller. Before running the controller component locally, it is necessary to configure kubeconfig in advance in the local environment (configured through the KUBECONFIG
environment variable or through the $HOME/.kube/config
file) and be able to access a Kubernetes cluster normally.
The Fluid CSI plugin cannot run locally to interact with the Kubernetes cluster. To run such components, it is necessary to first perform image construction, manually replace the corresponding image address of
charts/fluid/fluid/values.yaml
, and then deploy it to the Kubernetes cluster.
Ensure that go help is installed in the environment, and refer to the go installation manual for the specific installation process
$ dlv debug cmd/alluxio/main.go
If VSCode is used as the development environment, the Go plugin of VSCode can be directly installed and conduct local debugging.
Taking debugging the Alluxio Runtime Controller as an example, the Go code debugging task is defined in ./.vscode/launch.json
as follows:
{
"version": "0.2.0",
"configurations": [
{
"name": "Alluxio Runtime Controller",
"type": "go",
"request": "launch",
"mode": "debug",
"program": "cmd/alluxio/main.go",
"args": ["start", "--development=true", "--enable-leader-election"],
"env": {
"KUBECONFIG": "<path>/<to>/<kubeconfig>",
"ALLUXIO_RUNTIME_IMAGE_ENV": "alluxio/alluxio-dev:2.9.0",
"ALLUXIO_FUSE_IMAGE_ENV": "alluxio/alluxio-dev:2.9.0",
"DEFAULT_INIT_IMAGE_ENV": "fluidcloudnative/init-users:v0.8.0-5bb4677",
"MOUNT_ROOT": "/runtime-mnt",
"HOME": "<HOME_PATH>"
}
},
]
}
Proxy access to WebHook in the cluster to the local machine:
# 1. Install kt-connect(https://github.com/alibaba/kt-connect)
$ curl -OL https://github.com/alibaba/kt-connect/releases/download/v0.3.7/ktctl_0.3.7_Linux_x86_64.tar.gz
$ tar zxf ktctl_0.3.7_Linux_x86_64.tar.gz
$ mv ktctl /usr/local/bin/ktctl
$ ktctl --version
# 2. Proxy access to WebHook in the cluster to the local machine
$ ktctl exchange fluid-pod-admission-webhook --kubeconfig <path>/<to>/<kubeconfig> --namespace fluid-system --expose 9443
Set the debug task in ./.vscode/launch.json
:
{
"version": "0.2.0",
"configurations": [
{
"name": "Fluid Webhook",
"type": "go",
"request": "launch",
"mode": "debug",
"program": "cmd/webhook/main.go",
"args": ["start", "--development=true", "--full-go-profile=false", "--pprof-addr=:6060"],
"env": {
"TIME_TRACK_DEBUG": "true",
"MY_POD_NAMESPACE": "fluid-system"
}
},
]
}
For components such as Fluid Webhook and Fluid CSI plugins, remote debugging is usually the more commonly used method. Please ensure that go help is correctly installed on both the local machine and component images.
On remote host:
$ dlv debug --headless --listen ":12345" --log --api-version=2 cmd/alluxio/main.go
This will cause the remote host's debugging program to listen to the specified port (e.g. 12345)
On local machine:
$ dlv connect "<remote-addr>:12345" --api-version=2
Note: To debug remotely, make sure the specified port is not occupied and the firewall has been properly configured.