This module implements an even-driven file/batch based processing module as a microservice deployment using docker containers.
You can build this module with Maven:
mvn clean install
If you'd like to skip tests:
mvn clean install -Dtest=false -DfailIfNoTests=false
You should be able to run this locally using mvn, and it should work as expected. We highly recommend you test your services out locally before packaging as Docker containers:
mvn camel:run
You can then copy files to the target/placeorder
file location and watch the camel route get triggered and process the file. You should see an exception eventually because the camel route tries to place the data onto a A-MQ queue which would not be available at the time of running.
cp rider-auto-common/src/data/message1.xml rider-auto-file/target/placeorder/
You may also want to try the same thing using mvn exec:java
mvn exec:java
This will run the app's main file (from camel-boot) using the classpath from maven.
FIS includes a supported maven plugin for building Docker images from your project. Unless you're running on Linux, you'll need to install a guest VM that has Docker. We recommend the Red Hat Container Development Kit v2
For the maven plugin to work, it will need to be able to locate a Docker Daemon (ideally running in a guest-local VM). To do this, you'll want to have the following environment variables set:
export DOCKER_HOST=tcp://10.1.2.2:2376
export DOCKER_CERT_PATH=/path/to/cdk/.vagrant/machines/default/virtualbox/.docker
export DOCKER_TLS_VERIFY=1
Note the DOCKER_HOST
needs to point to the location of the docker daemon, and DOCKER_CERT_PATH
needs to point to the location of the cert for your docker daemon.
It would also be great if you had the native docker CLI tools installed on your Host machine (ie your Windows or Mac machine) so you can docker images
and docker ps
Before we build the docker image, we should specify whether we're going to use camel-boot (Plain Old Java Main with flat-classloader) or a Karaf osgi classloader. We recommend building using camel-boot. Basically what happens is we zip up the maven classpath and inject a little helper script to run the app using the classpath. This simplifies the deployment so we don't have to guess about which classpath we'll be using; it'll be exactly the same as when you ran locally.
To do this, run the following command:
$ mvn clean install -Pfabric8 -Phawtapp
(Skip tests can be done with adding -Dtest=false -DfailIfNoTests=false
)
After the build you should see the following in the target dir:
drwxr-xr-x 6 ceposta staff 204B Jan 26 13:15 classes
drwxr-xr-x 3 ceposta staff 102B Jan 26 13:15 fabric8
drwxr-xr-x 3 ceposta staff 102B Jan 26 13:15 generated-sources
drwxr-xr-x 3 ceposta staff 102B Jan 26 13:15 generated-test-sources
drwxr-xr-x 4 ceposta staff 136B Jan 26 13:15 hawt-app
drwxr-xr-x 3 ceposta staff 102B Jan 26 13:15 maven-status
-rw-r--r-- 1 ceposta staff 47M Jan 26 13:15 rider-auto-file-5.0-SNAPSHOT-app.tar.gz
-rw-r--r-- 1 ceposta staff 8.4K Jan 26 13:15 rider-auto-file-5.0-SNAPSHOT.jar
drwxr-xr-x 4 ceposta staff 136B Jan 26 13:15 test-classes
Notice the *.tar.gz file (can also generate zip file by setting the hawtapp-maven-plugin archiver to "zip"). In that zip/tar file, you'll find a completely packed up and ready to run app. For example if you unzip that file, you'll get the following contents:
drwxr-xr-x 3 ceposta staff 102B Jan 26 08:39 bin
drwxr-xr-x 200 ceposta staff 6.6K Jan 26 08:39 lib
From that folder, you can run:
./bin/run.sh
Which should bootstrap the application and run it as a standalone camel-boot app.
Now that you understand the camel-boot packaging using hawtapp-maven-plugin, we can build a docker container:
$ mvn clean install -Pfabric8 -Phawtapp docker:build
(Skip tests can be done with adding -Dtest=false -DfailIfNoTests=false
)
Note, we have to have a working docker daemon available as mentioned above.
You should end up with output similar to this:
[INFO] --- docker-maven-plugin:0.13.6:build (default-cli) @ rider-auto-file ---
[INFO] Copying files to /Users/ceposta/dev/sandbox/RedHatWorkshops/rider-auto-osgi/rider-auto-file/target/docker/fabric8/rider-auto-file/5.0-SNAPSHOT/build/maven
[INFO] Building tar: /Users/ceposta/dev/sandbox/RedHatWorkshops/rider-auto-osgi/rider-auto-file/target/docker/fabric8/rider-auto-file/5.0-SNAPSHOT/tmp/docker-build.tar
[INFO] DOCKER> [fabric8/rider-auto-file:5.0-SNAPSHOT] : Built image 8b0543d81d61
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 43.560 s
[INFO] Finished at: 2016-01-26T13:20:38-07:00
[INFO] Final Memory: 78M/1149M
[INFO] ------------------------------------------------------------------------
Now if you do a docker images
you should see your new docker image:
REPOSITORY TAG IMAGE ID CREATED VIRTUAL SIZE
fabric8/rider-auto-file 5.0-SNAPSHOT 8b0543d81d61 35 seconds ago 485 MB
You can even try to run you docker container as is and map the ports locally so you can see the service running within the docker image:
docker run -it --rm -v /home/vagrant/camel:/deployments/target/placeorder fabric8/rider-auto-file:5.0-SNAPSHOT
Note for that to work, we are mounting a location on the Vagrant VM (or the host Docker VM), so we will need to vagrant ssh
into the CDK box and copy a file to /home/vagrant/ camel to try it out
Yay! You now have your microservice packaged as a docker image ready to go. Let's take a look at what that looks like if you want to build the karaf-based microservice:
We can also convert existing karaf-based deployments over to the FIS deployment model. Just like with the camel-boot option (above) that packages the entire JVM together with its dependencies, we'll be doing that with Karaf. What this means is that your build will actually produce a completely independent Karaf assembly which can then be used to run your application. No more build your app and chuck it into a running Karaf and hope it resolves; now all OSGI resolution is done at build time and the resulting output is a fully baked Karaf distribution with your app inside it.
Run the following command to do this:
mvn clean install -Pfabric8 -Pkaraf-distro
(Skip tests can be done with adding -Dtest=false -DfailIfNoTests=false
)
That should produce output like this in /target
drwxr-xr-x 7 ceposta staff 238B Jan 31 15:49 assembly
drwxr-xr-x 5 ceposta staff 170B Jan 31 15:49 classes
drwxr-xr-x 3 ceposta staff 102B Jan 31 15:49 fabric8
drwxr-xr-x 3 ceposta staff 102B Jan 31 15:49 generated-test-sources
drwxr-xr-x 3 ceposta staff 102B Jan 31 15:49 maven-status
-rw-r--r-- 1 ceposta staff 5.7K Jan 31 15:49 rider-auto-file-5.0-SNAPSHOT.jar
-rw-r--r-- 1 ceposta staff 37M Jan 31 15:49 rider-auto-file-5.0-SNAPSHOT.tar.gz
-rw-r--r-- 1 ceposta staff 37M Jan 31 15:49 rider-auto-file-5.0-SNAPSHOT.zip
drwxr-xr-x 4 ceposta staff 136B Jan 31 15:49 test-classes
The .zip/.tar.gz file is the fully-baked karaf assembly. If you unzip it, it looks like a Karaf/Fuse distribution as you'd expect:
drwxr-xr-x 20 ceposta staff 680B Jan 26 08:58 bin
drwxr-xr-x 3 ceposta staff 102B Jan 26 08:58 data
drwxr-xr-x 45 ceposta staff 1.5K Jan 26 08:58 etc
drwxr-xr-x 13 ceposta staff 442B Jan 26 08:58 lib
drwxr-xr-x 7 ceposta staff 238B Jan 26 08:58 system
Can run ./bin/karaf
to boot up the karaf distro.
mvn clean install -Pfabric8 -Pkaraf-distro docker:build
(Skip tests can be done with adding -Dtest=false -DfailIfNoTests=false
)
Note, we have to have a working docker daemon available as mentioned above.
You should end up with output similar to this:
[INFO]
[INFO] --- docker-maven-plugin:0.13.6:build (default-cli) @ rider-auto-file ---
[INFO] Copying files to /Users/ceposta/dev/sandbox/RedHatWorkshops/rider-auto-osgi/rider-auto-file/target/docker/fabric8/rider-auto-file/5.0-SNAPSHOT/build/maven
[INFO] Building tar: /Users/ceposta/dev/sandbox/RedHatWorkshops/rider-auto-osgi/rider-auto-file/target/docker/fabric8/rider-auto-file/5.0-SNAPSHOT/tmp/docker-build.tar
[INFO] DOCKER> [fabric8/rider-auto-file:5.0-SNAPSHOT] : Built image fe32ffcad9e1
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 31.485 s
[INFO] Finished at: 2016-01-31T15:51:54-07:00
[INFO] Final Memory: 72M/853M
[INFO] ------------------------------------------------------------------------
Now if you do a docker images
you should see your new docker image:
REPOSITORY TAG IMAGE ID CREATED VIRTUAL SIZE
fabric8/rider-auto-file 5.0-SNAPSHOT fe32ffcad9e1 2 minutes ago 485 MB
You can even try to run you docker container as is and map the ports locally so you can see the service running within the docker image:
docker run -it --rm -v /home/vagrant/camel:/deployments/target/placeorder fabric8/rider-auto-file:5.0-SNAPSHOT
You should be able to copy files to /home/vagrant/camel and watch the logs of the container with
oc logs
To deploy into openshift, we need to generate the correct JSON manifest which includes all of our Services, Replication Controllers and Pods. We can do this a few different ways but for this module we use the fabric8 maven plugin. The plugin basically scans the list of maven properties and will generate a Kubernetes json or OpenShift template depending on the properties.
For example:
<fabric8.service.name>${project.artifactId}</fabric8.service.name>
<fabric8.service.headless>true</fabric8.service.headless>
<fabric8.metrics.scrape>true</fabric8.metrics.scrape>
<fabric8.metrics.port>9779</fabric8.metrics.port>
<fabric8.label.component>${project.artifactId}</fabric8.label.component>
<fabric8.label.container>java</fabric8.label.container>
<fabric8.label.group>rider-auto</fabric8.label.group>
<fabric8.iconRef>camel</fabric8.iconRef>
These maven properties will produce a JSON output when you build the project:
$ mvn clean install
The location of the kubernetes.json
file is in target/classes/kubernetes.json
However, we will want to customize this json file a little bit because we want to add volume and PersistentVolumeClaim information to be able to run in a kubernetes cluster. To do that, we're going to use the type-safe DSL for generating Kubernetes JSON using annotation processors. To enable the annotation processors, run the following mvn command (and examine the output of the kubernetes.json file, which should include the PersistentVolumeClaim and volume mount information:
mvn clean install -Pfabric8 -Pkube-generate
(Skip tests can be done with adding -Dtest=false -DfailIfNoTests=false
)
{
"apiVersion" : "v1",
"kind" : "Template",
"labels" : { },
"metadata" : {
"annotations" : {
"fabric8.rider-auto-file/iconUrl" : "https://cdn.rawgit.com/fabric8io/fabric8/master/fabric8-maven-plugin/src/main/resources/icons/camel.svg"
},
"labels" : { },
"name" : "rider-auto-file"
},
"objects" : [ {
"apiVersion" : "v1",
"kind" : "Service",
"metadata" : {
"annotations" : {
"prometheus.io/port" : "9779",
"prometheus.io/scrape" : "true"
},
"labels" : {
"container" : "java",
"component" : "rider-auto-file",
"provider" : "fabric8",
"project" : "rider-auto-file",
"version" : "5.0-SNAPSHOT",
"group" : "rider-auto"
},
"name" : "rider-auto-file"
},
"spec" : {
"clusterIP" : "None",
"deprecatedPublicIPs" : [ ],
"externalIPs" : [ ],
"ports" : [ {
"port" : 1
} ],
"selector" : {
"container" : "java",
"project" : "rider-auto-file",
"component" : "rider-auto-file",
"provider" : "fabric8",
"group" : "rider-auto"
}
}
}, {
"apiVersion" : "v1",
"kind" : "PersistentVolumeClaim",
"metadata" : {
"annotations" : { },
"labels" : {
"project" : "rider-auto-file",
"component" : "rider-auto-file",
"provider" : "fabric8",
"group" : "rider-auto"
},
"name" : "rider-auto-file-pvc"
},
"spec" : {
"accessModes" : [ "ReadWriteOnce" ],
"resources" : {
"limits" : { },
"requests" : {
"storage" : "100Ki"
}
}
}
}, {
"apiVersion" : "v1",
"kind" : "ReplicationController",
"metadata" : {
"annotations" : { },
"labels" : {
"container" : "java",
"component" : "rider-auto-file",
"provider" : "fabric8",
"project" : "rider-auto-file",
"version" : "5.0-SNAPSHOT",
"group" : "rider-auto"
},
"name" : "rider-auto-file"
},
"spec" : {
"replicas" : 1,
"selector" : {
"container" : "java",
"component" : "rider-auto-file",
"provider" : "fabric8",
"project" : "rider-auto-file",
"version" : "5.0-SNAPSHOT",
"group" : "rider-auto"
},
"template" : {
"metadata" : {
"annotations" : { },
"labels" : {
"container" : "java",
"component" : "rider-auto-file",
"provider" : "fabric8",
"project" : "rider-auto-file",
"version" : "5.0-SNAPSHOT",
"group" : "rider-auto"
}
},
"spec" : {
"containers" : [ {
"args" : [ ],
"command" : [ ],
"env" : [ {
"name" : "KUBERNETES_NAMESPACE",
"valueFrom" : {
"fieldRef" : {
"fieldPath" : "metadata.namespace"
}
}
} ],
"image" : "fabric8/rider-auto-file:5.0-SNAPSHOT",
"name" : "rider-auto-file",
"ports" : [ {
"containerPort" : 8778,
"name" : "jolokia"
} ],
"securityContext" : { },
"volumeMounts" : [ {
"mountPath" : "/deployments/target/placeorder",
"name" : "rider-auto-file-volume",
"readOnly" : false
} ]
} ],
"imagePullSecrets" : [ ],
"nodeSelector" : { },
"volumes" : [ {
"name" : "rider-auto-file-volume",
"persistentVolumeClaim" : {
"claimName" : "rider-auto-file-pvc",
"readOnly" : false
}
} ]
}
}
}
} ],
"parameters" : [ ]
}
The way Kubernetes works, if you want to store data to the file system in a persistent way, you can mount Volumes and use PersistentVolumeClaims to attach a pod to a volume. Above, we created the PersistentVolumeClaim using the type-safe DSL builders, but before we can spin up the app to use the PersistentVolumeClaim, we need to first establish a PersistentVolume. In general, only cluster admins or project admins will be able to administer the management of PersistentVolumes.
The PersistentVolume definition is located at src/main/fabric8/vagrant-pv.yaml and looks like this:
kind: PersistentVolume
apiVersion: v1
metadata:
name: rider-auto-file-pv
labels:
type: local
spec:
capacity:
storage: 100Ki
accessModes:
- ReadWriteOnce
persistentVolumeReclaimPolicy: Recycle
hostPath:
path: "/opt/camel"
So as an cluster or project admin user, you can run the following to create the PersistentVolume:
oc create -f vagrant-pv.yaml
Note: as a user of the CDK, you can achieve cluster-admin rights by using the config file in /var/lib/origin/openshift.local.config/master/admin.kubeconfig. If you copy this to your ~/.kube/config file or set the location to it in $KUBECONFIG environment variable, you should be granted full cluster-admin rights and should be able to create PVs. It is best to first login to the vagrant from your host machine:
vagrant ssh
Then sudo to root
sudo -s
Then login to OpenShift using the "system:admin" superuser on a dedicated admin interface (10.0.2.15):
oc login -u system:admin https://10.0.2.15:8443
Now you can create the PersistentVolume by running the following command:
oc create -f vagrant-pv.yaml
After you've created the PersistentVolume, you should be able to verify:
[root@localhost ~]# oc get pv
NAME LABELS CAPACITY ACCESSMODES STATUS CLAIM REASON AGE
rider-auto-file-pv type=local 100Ki RWO Available 5d
Now we can deploy our app and expect the Claim to be satisfied. Note, in this example, we're using a HostVolume, but in practice that is not advisable.
We can take that json and "apply" it to a running OpenShift installation. Note that for this plugin to work, you must
already be logged into openshift with oc login
and have the following environment variable set:
export KUBERNETES_MASTER=https://10.1.2.2:8443
export KUBERNETES_DOMAIN=
export KUBERNETES_TRUST_CERT=true
Then you should be able to run the following maven command to deploy:
$ mvn clean install -Pfabric8 -Pkube-generate fabric8:apply
Note for this to work, you must have run the docker:build
previously. Or you can combine all of them:
$ mvn clean install -Pfabric8 -Phawtapp -Pkube-generate docker:build fabric8:apply
(Skip tests can be done with adding -Dtest=false -DfailIfNoTests=false
)
This command assumes everything is running locally (like on the CDK). Otherwise, if deploying to a centralized openshift
installation, you can build the docker image locally, then do docker:push
to get the image up to OpenShift and then run
the fabric8:apply
command to deploy the app.
This example uses JBoss A-MQ, so we need to have that running in the same project/namespace as the rider-auto apps (including this module). To deploy AMQ, follow the instructions from the xPaaS AMQ documentation. On the CDK, you can do this:
oc create -f https://raw.githubusercontent.com/openshift/openshift-ansible/master/roles/openshift_examples/files/examples/v1.1/xpaas-templates/amq62-basic.json oc process amq62-basic -v APPLICATION_NAME=broker -v MQ_USERNAME=admin -v MQ_PASSWORD=admin
Or you can use the template i've included in the root of this project:
oc create -f amq.json
service "broker-amq-tcp" created
deploymentconfig "broker-amq" created
Note that the user name and password need to be admin/admin
as that's what the rider-auto-osgi project expects.