Re: Docker agent build script wrapper
>
> We have reached acceptable solution for the problem. It is loading groovy
> script with black magic around closure. Attaching code for the community
> benefit.
>
It boils down to construct in the loaded groovy module:
def container(myImageName, Closure body){
docker.image(myImageName).inside(dockerArgs(), body)
}
And then using it passing closure body:
def myImage = 'php:latest'
def modelId = drun.modelId()
echo modelId
drun.container(myImage) {
stage ('setup'){
sh 'id'
echo 'value = ' + modelId
}
stage('stage 1'){
sh '/usr/local/bin/php -version'
}
}
There is also custom step returning map of properties from the environment
implemented as Java class:
public class DrunPropertiesStep extends Step {
...
public static class Execution extends SynchronousStepExecution {
private static final long serialVersionUID = 1;
Execution(StepContext context) {
super(context);
}
@Override
protected Object run() throws Exception {
Run run = getContext().get(Run.class);
//todo: implement
JSONObject json = Utils.getJson(run.getRootDir().getPath(),
run.getLogReader());
Map map = new HashMap();
Iterator iter = json.keys();
while( iter.hasNext() ) {
String key = iter.next().toString();
map.put(key, json.getString(key));
}
return map;
}
}
...
}
Hope it will help to prevent someone from banging a head against the
keyboard :)
Thanks
Alex
--
You received this message because you are subscribed to the Google Groups
"Jenkins Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to jenkinsci-dev+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/jenkinsci-dev/3cd3005a-9a64-48a4-b67b-2cae076d0e93%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
drun.groovy
Description: Binary data
Jenkinsfile
Description: Binary data
Re: Docker agent build script wrapper
Hi,
next release of docker-plugin is going to introduce "dockerNode" keyword in
pipeline, so you can create an agent from a docker image:
dockerNode(dockerHost: 'unix:///var/run/docker.sock', image:
'jenkins/slave', remoteFs: '/home/jenkins') { sh 'echo "hello there"' }
this is still under development and will first be an experimental feature
in the meantime, you can rely on docker-pipeline-plugin and docker.inside
for an equivalent.
Generally speaking, if you target multiple environments, you should not
declare a top level node, but have one per stage.
2017-11-27 11:12 GMT+01:00 Oleksandr Rudak :
> Hi,
>
> We are building scientific model running and testing environment product.
> We use Jenkins as a core building, integration and model running tool. Now
> our users are scientist and have little technological background so we look
> to fold our current Jenkinsfile into something more compact and
> transparent. We've tried few things. SimpleBuildStep, which is too basic to
> handle agent/docker stuff. Step is too complicated as it involved
> distributed execution parallelism. Also we tried 'load' to load Jenkinsfile
> generated in the working directory. All ways have pros/cons but none
> provides in full what we try to accomplish.
>
> Any ideas how to do such wrapping? I think about custom `agent` which
> would provide docker context and will imply the rest of the configuration
> and hide it from user. Possibly some workflow extension or something like
> this. Any thoughts are welcome.
>
> Below is the anticipated Jenkinsfile how it would ideally look. Attached
> is the original Jenkins with detailed steps we want to cloak from user.
>
> node {
> stage ('setup') {
> setupEnv (dockerImage:'dockerImage123')
> }
> stage ('train') {
> runNotebook (notebookPath:'notebookPath212')
> // { prepare, run, archive, legion }
> // runNotebook (notebookPath:'notebookPath213')
> }
> stage ('deploy') {
> deployModel ()
> }
> stage ('test') {
> testModel ()
> }
> }
>
>
>
> Thanks
> Alex
>
> --
> You received this message because you are subscribed to the Google Groups
> "Jenkins Developers" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to jenkinsci-dev+unsubscr...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/
> msgid/jenkinsci-dev/5a2441d6-72ef-45f7-a415-356510c8a9cc%
> 40googlegroups.com
>
> .
> For more options, visit https://groups.google.com/d/optout.
>
--
You received this message because you are subscribed to the Google Groups
"Jenkins Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to jenkinsci-dev+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/jenkinsci-dev/CANMVJzmLW_c67%3DrO7CLk7k_L5FZeZFiw33SGQh2X36Ty-rovDg%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.
Docker agent build script wrapper
Hi,
We are building scientific model running and testing environment product.
We use Jenkins as a core building, integration and model running tool. Now
our users are scientist and have little technological background so we look
to fold our current Jenkinsfile into something more compact and
transparent. We've tried few things. SimpleBuildStep, which is too basic to
handle agent/docker stuff. Step is too complicated as it involved
distributed execution parallelism. Also we tried 'load' to load Jenkinsfile
generated in the working directory. All ways have pros/cons but none
provides in full what we try to accomplish.
Any ideas how to do such wrapping? I think about custom `agent` which would
provide docker context and will imply the rest of the configuration and
hide it from user. Possibly some workflow extension or something like this.
Any thoughts are welcome.
Below is the anticipated Jenkinsfile how it would ideally look. Attached is
the original Jenkins with detailed steps we want to cloak from user.
node {
stage ('setup') {
setupEnv (dockerImage:'dockerImage123')
}
stage ('train') {
runNotebook (notebookPath:'notebookPath212')
// { prepare, run, archive, legion }
// runNotebook (notebookPath:'notebookPath213')
}
stage ('deploy') {
deployModel ()
}
stage ('test') {
testModel ()
}
}
Thanks
Alex
--
You received this message because you are subscribed to the Google Groups
"Jenkins Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to jenkinsci-dev+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/jenkinsci-dev/5a2441d6-72ef-45f7-a415-356510c8a9cc%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
pipeline {
agent {
docker {
image 'drun.kharlamov.biz/drun/base-python-image:latest'
args '-v drun:/drun -u 0:0 --network=drun_drun_root -e
"DOCKER_TLS_VERIFY=1" -e "DOCKER_HOST=${DOCKER_HOST}" -e
"DOCKER_CERT_PATH=${DOCKER_CERT_PATH}" -e
"MODEL_SERVER_URL=${DRUN_MODEL_SERVER_URL}"'
}
}
stages {
stage('clone'){
steps {
sh '''
pip install -i https://drun.kharlamov.biz/pypi/
drun
'''
}
}
stage('run notebook'){
steps {
sh '''
jupyter nbconvert --execute
"Digit-Recognition/Recognizing digits.ipynb"
cp "Digit-Recognition/Recognizing digits.html"
notebook.html
mkdir -p release-models
cp -rf /drun/* release-models/
'''
}
}
stage('archive notebook artifacts'){
steps {
archiveArtifacts
'release-models/image_recognize.model'
archiveArtifacts 'Digit-Recognition/*.html'
archiveArtifacts 'notebook.html'
}
}
stage('deploy to legion'){
steps {
sh '''
legion build
release-models/image_recognize.model
legion deploy --model-id recognize_digits
'''
sleep time: 20, unit: 'SECONDS'
}
}
stage('run tests'){
steps {
sh '''
cd "Digit-Recognition/tests"
nosetests --with-xunit
'''
junit 'Digit-Recognition/tests/nosetests.xml'
}
}
}
}
