Skip to content

Latest commit

 

History

History
605 lines (434 loc) · 24.3 KB

2-dev.md

File metadata and controls

605 lines (434 loc) · 24.3 KB
Raw

Basic CloudStack Development

CloudStack Development 101

Overview video: https://s3-eu-west-1.amazonaws.com/shapeblue-engineering-videos/hackerbook/2-dev/2-dev-overview.mp4

Guided walkthrough video: https://s3-eu-west-1.amazonaws.com/shapeblue-engineering-videos/hackerbook/2-dev/2-dev-guided-walkthrough.mp4

From an application setup and functioning perspective, a typical CloudStack deployment consists of the following:

  • CloudStack Management server(s)
  • CloudStack Agent (on KVM)
  • CloudStack Usage server(s)

The CloudStack management server is a monolith Java application that embeds the control plane, orchestrator and overall cloud controller.

The management server (sometimes written as mgmt server) has three types of APIs:

  • Platform API (REST-like or query-based API for end users and admins)
  • Agent API (ServerResource based json/API that uses command-answer pattern handled by an implementation that can talk to a hardware resource)
  • Plugin API (Set of Java interfaces and APIs to allow extension and modification of CloudStack)

The management server monolith has various layers:

  • Presentation or API: the layer that implements and handles REST-like or query based APIs.
  • Service/Business/Orchestration layer: the layer that usually handles API requests, manages business entities and participates in resource control.
  • DB/Data access: the data access layer implements set of building blocks to talk to the database (MySQL).
  • Kernel: Provides building blocks, polling, IPC, message bus, implements orchestration and controller for compute, network, storage, security etc.
  • Agent/Cluster management: the layer that handles distributed system of mgmt server nodes and agents, and handles RPC mechanisms.

Recommended Reading:

Old videos:

Setting up Development Environment

Video: https://s3-eu-west-1.amazonaws.com/shapeblue-engineering-videos/hackerbook/2-dev/2-dev-chapter.mp4

The recommended development environment is Linux based, in this course Ubuntu Linux 20.04+ is preferred. Run the following to install packages required for CloudStack development on Ubuntu: (see first chapter on software installation)

$ sudo apt-get install openjdk-11-jdk maven mysql-server mysql-client bzip2 nfs-common uuid-runtime python3-setuptools ipmitool genisoimage nfs-kernel-server quota

Older CloudStack versions may require older jdk/jre version, therefore setup, install and learn to use jenv: http://www.jenv.be and also run jenv enable-plugin maven.

The recommended setup is to run MySQL and NFS servers locally on your laptop and the hypervisor in a VM or an external host. The CloudStack management and usage server can be run using maven or via an IDE.

Tip: get the latest maven from https://maven.apache.org/download.cgi

Getting the source code

Apache CloudStack source code can be cloned from the following remotes:

Create a Gitbub account in case you do not have one already: https://github.com

You may generate a public SSH key (if do not already have one at: ~/.ssh/id_rsa.pub):

    ssh-keygen -t rsa -b 4096 -C "[email protected]"
    eval "$(ssh-agent -s)"
    ssh-add ~/.ssh/id_rsa

Add your SSH key to you Github account:

  • Go to your Github account
  • Click on Settings and select SSH and GPG keys
  • Click on New SSH key
  • Add a title and paste the content of your key
  • Click Add SSH key

You may create a personal workspace and clone the repository, for example:

mkdir -p ~/lab/
cd ~/lab/
git clone https://github.com/apache/cloudstack.git

The recommended directory structure may look something like:

        ~/lab/
        ├── cloudstack
        └── cloudmonkey
        └── shapeblue
            └── ... private repositories ...
        └── ... other projects ...

Reference and reading resources:

Development tools to learn:

  • git, tig
  • maven
  • IntelliJ IDEA IDE, vim
  • java, javac, jps, jstack, jmap
  • IntelliJ remote debugger, VisualVM, MAT

Setup IDE

Setup IntelliJ IDEA (recommended) or any IDE of your choice. Get IntelliJ IDEA community edition from:

https://www.jetbrains.com/idea/download/#section=linux

Or you can install then using snap: (preferred)

sudo snap install intellij-idea-community --classic

Start the IDE, configure as needed and import CloudStack as:

  • Click on import project
  • Select the cloned cloudstack directory
  • Select Maven as build system
  • Select options suitably and import!

You may configure IDEA per your preference or use this settings jar which you can import in IDEA as File > Import Settings > select jar file.

Setup MySQL Server

After installing MySQL server, configure the following settings in its config file such as at /etc/mysql/mysql.conf.d/mysqld.cnf and restart mysql-server:

[mysqld]

sql-mode="STRICT_TRANS_TABLES,NO_ENGINE_SUBSTITUTION,ERROR_FOR_DIVISION_BY_ZERO,NO_ZERO_DATE,NO_ZERO_IN_DATE,NO_ENGINE_SUBSTITUTION"
server_id = 1
innodb_rollback_on_timeout=1
innodb_lock_wait_timeout=600
max_connections=1000
log-bin=mysql-bin
binlog-format = 'ROW'
# add the following for MySQL 8.x
default-authentication-plugin=mysql_native_password

Tip: ensure that your mysql server only listens on 127.0.0.1 and reset the mysql root password to blank to get CloudStack db-deployment using mvn work out of the box, run sudo mysql -u root -e "ALTER USER 'root'@'localhost' IDENTIFIED WITH mysql_native_password BY ''" (tested with mysql 5.7+).

Setup NFS storage

After installing nfs server, configure the exports:

echo "/export  *(rw,async,no_root_squash,no_subtree_check)" > /etc/exports
mkdir -p /export/testing/primary /export/testing/secondary

Tip: You may want to have separate secondary storage for each version of CloudStack. Rename and add more directories to the /export/testing path as and when required.

The following is one way to seed a systemvmtemplate, for example for the 4.15:

wget http://packages.shapeblue.com/systemvmtemplate/4.15/systemvmtemplate-4.15.0-kvm.qcow2.bz2
./scripts/storage/secondary/cloud-install-sys-tmplt \
      -m /export/testing/secondary -f systemvmtemplate-4.15.0-kvm.qcow2.bz2 \
      -h kvm -o localhost -r cloud -d cloud

Notes:

  • Please check the branch you're building/working against, and use the suitable systemvmtemplate version for your branch. For master, use the latest systemvmtemplate or build your own using packer, in the source see tools/appliance/README.md for details.
  • Deploy the systemvm template after deploying the CloudStack database, refer to the Building CloudStack section below.

Configure Environment

Put the following in your sudoers file using visudo etc. to allow processes owned by your user (such as the CloudStack management server) to execute some privileged commands:

Cmnd_Alias CLOUDSTACK = /bin/mkdir, /bin/mount, /bin/umount, /bin/cp, /bin/chmod, /usr/bin/keytool, /bin/keytool

Defaults:username !requiretty

username ALL=(ALL) NOPASSWD:CLOUDSTACK

Tip: replace username with your Linux username.

Building CloudStack

Noredist CloudStack builds requires additional jars that may be installed from:

https://github.com/shapeblue/cloudstack-nonoss

Clone the above repository and execute the install script to install the noredist jar dependencies:

$ cd /path/to/cloudstack-nonoss/repo
$ bash -x install-non-oss.sh

To build CloudStack with noredist (this include vmware plugins etc):

    $ cd /path/to/cloudstack
    $ mvn clean install -Dnoredist -P developer,systemvm

Deploy CloudStack database using:

$ mvn -Pdeveloper -pl developer -Ddeploydb

Run management server using:

$ mvn -pl :cloud-client-ui jetty:run -Dnoredist -Djava.net.preferIPv4Stack=true

To deploy an environment, you can use the deploy datacenter script as:

$ python3 tools/marvin/marvin/deployDataCenter.py -i /path/to/config.cfg

Note : The sample cfg files are present in setup/dev

Note: Use pip3 to install/upgrade any dependencies for Marvin

$ pip3 install --upgrade tools/marvin/dist/Marvin-*.tar.gz

Example of how to run a marvin based integration test: (change parameters suitably)

$ nosetests --with-xunit --xunit-file=results.xml --with-marvin --marvin-config=/path/to/config.cfg -s -a tags=advanced --hypervisor=KVM test/integration/smoke/test_vm_life_cycle.py

When needed, the usage server can be started using:

$ mvn -P usage -Drun -Dpid=$$ -pl usage

Note: due to bug in surefire plugin you may need to use -Djdk.net.URLClassPath.disableClassPathURLCheck=true.

Build tips:

  • In case the build fails on Ubuntu 22.04, please make sure there is correct alias for python
    $ sudo apt install python-is-python3

or create an alias. Place this into ~/.bashrc or ~/.zshrc file:

    $ alias python=python3

  • For an iterative styled development and code building, you may use the mvn -pl or --projects flag to which you can pass comma separate list of maven projects or paths you've changed and the client (which builds a fatjar based on all other projects), for example if you only changed api and vmware:

    $ mvn clean install -Dnoredist -P developer,systemvm -pl api,plugins/hypervisor/vmware,client

  • You may skip unit tests and even build with parallel threads:

    $ mvn clean install -Dnoredist -P developer,systemvm -DskipTests -T8

  • You will need to build and install/update marvin library every time you change a major CloudStack branch and/or if you add/remove/modify CloudStack APIs.

  • To bring up the Cloudstack Ui

    $ cd /path/to/cloudstack/ui
    $ sudo apt install nodejs npm
    $ npm install
    $ npm run serve

Make sure to set CS_URL=http://localhost:8080/client on .env.local file on ui.

You should be able to run the management server on http://localhost:5050

Testing CloudStack

Unit Testing

Unit tests in CloudStack are generally written with JUnit4 that also use mockito, powermock and sometimes wiremock. You may learn more about JUnit4 and usage of other libraries using existing unit tests (Ctrl+Shift+t in IntelliJ to see an existing class's unit test) or by using following references:

Functional Testing

Functional or integration tests for CloudStack are written in Python 2.7 using the Marvin test library. At the core of it, these are basically Python unittest tests where the following test-probes are available:

  • API client: run and get result of CloudStack APIs
  • DB client: query MySQL database
  • SSH client: allow remote access into a host endpoint/port

Typical integration tests are run using nose (a test runner), and an integration test basically is a Python class that extends cloudstackTestCase which are based on unittest.case.TestCase. The nature and mechanism of how these tests run, make them integration or functional tests.

You may use following references to know more about using Python unittest framework:

The functional testing exercise will cover in much detail how to write marvin based integration tests and use the utilities, probes and other building blocks of the Marvin library. As an example, you may look at the test/integration/smoke/test_dynamicroles.py marvin test.

For developing and debuging Marvin tests you may want to use a python IDE, for example PyCharm (https://www.jetbrains.com/pycharm/download/) and its community edition.

Once installed PyCharm, you'll need to configure the project interpreter into its preferences. When the interpreter is set, Marvin lib shouold appear in the available packages

In order to execute a marvin test you'll need to setup a Run Configuration

  • In PyCharm, click Run -> Edit Configurations -> add new python test configuration (nosetest)
  • Then add the following line into Additional Configurations --with-marvin --marvin-config=[path-to-config-file] -s -a tags=advanced --hypervisor=[xenserver|kvm|vmware|simulator] --zone=[zone-id]
  • Target -> Script Path and select the particular test file you want to execute for example: test/integration/smoke/test_dynamicroles.py Now you should be able to run/debug any Marvin test.

Simulator Based Development

CloudStack has a mocked hypervisor called simulator that may be used for development of presentation, service, db and orchestration layer features.

You can use the simulator flag to build the simulator hypervisor plugin as:

$ mvn clean install -Dsimulator -P developer,systemvm

Note in addition to deploying database the following must be run:

$ mvn -Pdeveloper -pl developer -Ddeploydb
$ mvn -Pdeveloper -pl developer -Ddeploydb-simulator

Run the management server using the simulator flag as well:

$ mvn -pl :cloud-client-ui jetty:run -Djava.net.preferIPv4Stack=true -Dsimulator

Simulator based environment can be deployed using:

$ pip3 install --upgrade tools/marvin/dist/Marvin-*.tar.gz
$ python3 tools/marvin/marvin/deployDataCenter.py -i setup/dev/advanced.cfg

MBX Based Development

mbx (MonkeyBox) is a VM appliance that runs a hypervisor such as KVM/VMware/XenServer in a Intel-VTx/AMD-v enabled VM on KVM (your laptop). Follow the MonkeyBox project for details: https://github.com/shapeblue/mbx

Video - https://www.youtube.com/watch?v=a8xuWHrzw88

Old Video for older monkeybox tool: (please refer to the above repo for latest mbx usage) https://s3-eu-west-1.amazonaws.com/shapeblue-engineering-videos/hackerbook/2-dev/2-dev-monkeybox.mp4

Debugging CloudStack

To debug any java process started by maven, you can export the following in your shell (or include this by default in zshrc or bashrc):

export MAVEN_OPTS="$MAVEN_OPTS -Xdebug -Xrunjdwp:transport=dt_socket,address=8787,server=y,suspend=n"

For installed CloudStack management server, you can change its JAVA_DEBUG in the /etc/default/cloudstack-management (see PR).

To remote-debug the KVM agent, put the following in /etc/default/cloudstack-agent in your monkeybox and restart cloudstack-agent:

JAVA=/usr/bin/java -Xdebug -Xrunjdwp:transport=dt_socket,address=8787,server=y,suspend=n

This will then allow you to attach a remote debugger on port 8787 (or any other port you may have configured).

For reference, please keep in mind the various ports used by CloudStack:

  • 8080: Authenticated API service
  • 8096: Unauthenticated API service
  • 9090: Cloudstack Management server cluster
  • 45219: JMX console
  • 8250: Management server port for agents
  • 3922: SSH port for systemvms
  • 3306: MySQL server
  • 22: KVM, XenServer/XAPI
  • 443: XenServer, vCenter
  • 53: DNS
  • 111/2049: NFS service/communication port, secondary storage VM
  • 860/3260: iSCSI communication port for iSCSI software connector
  • 7080: AWS API server (deprecated in recent versions)

Using Logs

For a typical CloudStack installation, logs are found per service as follows:

  • cloudstack-management: dir:/var/log/cloudstack/management/
  • cloudstack-usage: dir:/var/log/cloudstack/usage/
  • cloudstack-agent: dir:/var/log/cloudstack/agent/

However, when management server is launched using maven the logs will be in the root directory of your source directory:

  • vmops.log: the management server log
  • api.log: the API log
  • usage.log: the usage server log

Start the management server using mvn and try to read, follow and understand what may be happening when the management server starts. For example, the first thing you'll notice is that it:

  • Loads module context (spring-bootstrap-context.xml) and creates a hierarchy of modules to load
  • It starts loading modules in a certain order and starts integrity checks
  • It configures various CloudStack components, instantiates registeries etc.
  • Discovers API, starts various CloudStack components
  • Finally the management server API service is available at port 8080 (default) and log statement like the following are seen:
2018-10-27 00:53:06,417 INFO  [c.c.c.ClusterManagerImpl] (Cluster-Heartbeat-1:ctx-4108c094) (logid:505420fb) We are good, no orphan management server msid in host table is found
2018-10-27 00:53:06,420 INFO  [c.c.c.ClusterManagerImpl] (Cluster-Heartbeat-1:ctx-4108c094) (logid:505420fb) No inactive management server node found
2018-10-27 00:53:06,437 DEBUG [c.c.c.ClusterManagerImpl] (Cluster-Heartbeat-1:ctx-4108c094) (logid:505420fb) Detected management node joined, id:1, nodeIP:127.0.0.1

Tip: Several CloudStack operations are scheduled and executed by its job framework which gives such operations a unique job ID such as job-123 and this makes it easier to grep and investigate logs of a passed/failed job using the management server logs of a multi-tenant (huge) CloudStack deployment.

Challenge: Deploy a VM (either using Simulator or KVM/monkeybox) and tail through the logs and try to read and make sense of various steps that were performed between the API request to deploy a VM was initiated and when the VM came online.

Using IDE

With the Java process (management server or agent) launched with above mentioned flags, in IntelliJ IDEA go to Run > Debug... > Edit configuration > Add > Remote and configure it suitably. Put breakpoint in code, start the debugger, and wait for code execution such as an API request to reach the breakpoint to debug.

Challenge: Add a breakpoint at the execute() of the DeployVMCmdByAdmin class and step through the code to explore what gets executed when CloudStack deploys a VM. Reference you findings against what you learnt by going through the logs.

Instrumentation

Several instrumentation and debugging tools exists that may be used to debug a general Java/JVM application.

To list Java processes on a host:

jps -l  # or ps aux | grep java

To get the thread dump of a process:

jstack -l <pid>

To get the heap dump of a process:

jmap -dump:format=b,file=heap.bin <pid>

Few popular instrumentation tools:

  1. Visual VM

Debugging remotely:

- Setup socks proxy to the remote server: ssh -v -D 9696 root@<path>
- Configure VisualVM to use this as SOCKS proxy in Tools->Options->Network
- Run CloudStack management server with option: -Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.port=1099 -Dcom.sun.management.jmxremote.ssl=false -Dcom.sun.management.jmxremote.authenticate=false
- Add remote host IP and add JMX connection on it on port 1099

Case study: apache/cloudstack#2314

  1. Eclipse MAT

Case study: apache/cloudstack#1729

Other notable mentions:

CloudStack Packaging

CloudStack is typically packages and shipped as a deb or rpm repository. The docker container images from https://hub.docker.com/u/shapeblue can be used to build CloudStack for CentOS (rpm) and Ubuntu (rpm).

The pre-requisite is that all the build and runtime dependencies are installed on a build system (CentOS or Debian based) along with any nonoss dependencies.

Building on Debian/Ubuntu:

# Option 1: Building using commands
cd /root/of/the/cloudstack/repo
dpkg-buildpackage -uc -us -b

# Option 2: Building using script
cd /root/of/the/cloudstack/repo
bash -x packaging/build-deb.sh  # run with -h for help

Building on CentOS:

# For el6/centos6
cd /root/of/the/cloudstack/repo
bash -x packaging/package.sh -p noredist -d centos63  # run with -h for help

# For el7/centos7
cd /root/of/the/cloudstack/repo
bash -x packaging/package.sh -p noredist -o rhel7 -d centos7  # run with -h for help

Contributing to CloudStack

For bug reporting create an issue: https://github.com/apache/cloudstack/issues

For any bugfix or improvement change(s) send a pull request: https://github.com/apache/cloudstack/pulls

For feature submission the typical process is as follows:

  • Write a high level functional specification (FS).
  • Start a discussion on dev@ mailing list with reference to the FS and/or any issues. You can use [DISCUSS] in the email subject. Continue the discussion and engage with the community if anyone posts any feedback or has a question.
  • Complete the feature, send a pull request (PR).
  • Participate in code review, iterate implementation, request committers for review and merging. Typically every PR will be reviewed and (regression) tested. It is expected feature/bugfix PRs to have unit and integration tests.
  • Send documentation PR.

Case Study: Dynamic Roles feature

Basic Development Topics

# Topic Est
#1 Functional Spec 8 hours
#2 API Development 16 hours
#3 Service Development 16 hours
#4 DB Development 16 hours
#5 Pluggable Framework and Plugin development 8 hours
#6 IPC/RPC 8 hours
#7 Usage Development 8 hours
#8 UI Development 8 hours
#9 Functional Testing 8 hours
#10 Packaging 4 hours

Tip: Do api, service and db exercises together.