Skip to content

mvk13ogb/qwait

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

QWait

Build Status

This is QWait, the next generation queuing system for KTH CSC.

Trying out the application

Docker

The application is available in the public Docker registry. Simply run:

docker run -p 8080:8080 dflemstr/qwait

This is all you have to do, provided that you have Docker installed.

Manually

If you want to try out QWait, simply do this:

  • Install Java JDK version 7 or later.
  • Install Apache Maven version 3 or later.
  • Enter the qwait directory and type mvn jetty:run to test the application. You can reach it at http://localhost:8080/.

Note that unless you have configured LDAP correctly, logged-in users will have names that are not very readable, because the real names of users are fetched via LDAP queries.

It is also tricky to do anything with this test version of the application, because you will not be an administrator, and only administrators can add queues. It is recommended that you configure a database such as PostgreSQL as outlined below, so that you can configure the initial administrator via the database.

Deployment

QWait can be deployed in various ways; either continuously via a CI/CD system, or manually.

Continuous deployment

This repository is being continuously deployed. This allows for rapid development since no manual intervention is required while deploying.

The configuration for this system is in the .travis.yml file in the repository.

The automatic deployment process works like this:

  • A commit is made on the master branch of this repository.

  • Travis CI picks up the new commit and starts building it. One such build looked like this. Travis runs the following commands in order:

    # Fetch the source code
    git clone --depth=50 --branch=master git://github.com/mvk13ogb/qwait.git mvk13ogb/qwait
    cd mvk13ogb/qwait
    
    # Go to the particular commit we want to build (to avoid a race condition
    # where someone might have committed something else in the mean time)
    git checkout -qf <commit id>
    
    # Download dependencies & build everything
    mvn install -DskipTests=true -B -V
    
    # Run unit tests and integration tests
    mvn verify
    
  • If and only if the exit code of mvn verify is 0 will the next step be performed. Then, Travis will run the following script:

    if [[ "$TRAVIS_PULL_REQUEST" == false ]] && [[ "$TRAVIS_BRANCH" == master ]]
    then
        mvn -B package cargo:redeploy -DskipTests=true
    fi
    

    The additional check is there to ensure that the mvn command only runs if it is the master branch that is being built. The build configuration is used to build pull requests as well, and pull requests should not be deployed.

    As can be guessed from the command, it will package it as a .war file and deploy the web application using the Codehaus Cargo system. The configuration for Cargo is stored in the pom.xml file in the root of the repository. It might look something like this:

    <plugin>
        <groupId>org.codehaus.cargo</groupId>
        <artifactId>cargo-maven2-plugin</artifactId>
        <version>1.4.7</version>
        <configuration>
            <container>
                <containerId>jetty9x</containerId>
                <type>remote</type>
            </container>
    
            <configuration>
                <type>runtime</type>
                <properties>
                    <cargo.hostname>qwait.csc.kth.se</cargo.hostname>
                    <cargo.servlet.port>8080</cargo.servlet.port>
                    <cargo.remote.username>admin</cargo.remote.username>
                    <cargo.remote.password>${env.CARGO_PASSWORD}</cargo.remote.password>
                </properties>
            </configuration>
    
            <deployer>
                <type>remote</type>
            </deployer>
    
            <deployables>
                <deployable>
                    <groupId>${project.groupId}</groupId>
                    <artifactId>${project.artifactId}</artifactId>
                    <type>war</type>
                    <properties>
                        <context>/</context>
                    </properties>
                </deployable>
            </deployables>
        </configuration>
    </plugin>
    

    The configuration says that:

    1. The application should be deployed to a remote Jetty 9.x container.
    2. The container is located at http://qwait.csc.kth.se listening on port 8080.
    3. To authenticate with Cargo, the username admin is used. The password is stored in the CARGO_PASSWORD environment variable. Travis knows about this password because of the env[0].secure line in the .travis.yml file in the repository, which contains encrypted instructions telling Travis to export the CARGO_PASSWORD environment variable with the right contents.
    4. The deployer is also running remotely.
    5. The artifacts to be deployed include our current project in the POM, which should be deployed to the / (root) context on the server.
  • On the remote server, Jetty 9.x is installed, with a webapp called the cargo-jetty-7-and-onwards-deployer. It authenticates the deployment attempt and receives the .war file. The file is renamed to ROOT.war (to put it in the / context which has the special name ROOT) and saved in the webapps folder of the Jetty container, where Jetty picks it up and starts serving it.

Manual deployment

The webapp can be manually deployed like so:

  • Run mvn clean verify package in the source repository.
  • Copy the resulting .war file from the target/ directory onto the server machine into the webapps/ directory of Jetty, for example into /opt/jetty/webapps/. The name of the file will decide the context path, so naming the file foo.war will deploy it to http://server.se/foo/. The special name ROOT.war will deploy the webapp to http://server.se/.

Caveats

Deploying new versions of a web application without restarting Jetty for a long time might yield Permanent Generation memory issues in Jetty's JVM. The reason for this is that the web application's classes are reloaded each time the application is deployed, and sometimes old classes are not successfully evicted from memory. It is therefore recommended to restart Jetty after each deploy, but this is not necessary most of the time.

Setting it up

The easy way

Simply checking out the repository and running mvn jetty:run is enough to get a fully functional web application up and running, listening to port 8080. See point 4 below on how to forward port 8080.

The hard way

To set up deployment, a few configuration steps are necessary:

  1. Download and install a Java 7 JRE. On Ubuntu, this is simply done with:

    sudo aptitude install openjdk-7-jre
    

    Then, switch to that Java installation by default with:

    sudo update-java-alternatives -s java-1.7.0-openjdk-amd64
    

    To confirm that everything worked, try this and confirm that the version is 1.7.0:

    $ java -version
    java version "1.7.0_55"
    OpenJDK Runtime Environment (IcedTea 2.4.7) (7u55-2.4.7-1ubuntu1~0.12.04.2)
    OpenJDK 64-Bit Server VM (build 24.51-b03, mixed mode)
    
  2. Download and install a recent version of Jetty or some other Servlet container. The webapp has so far been tested with Jetty, so your mileage might vary. The container must support the Servlet v3.0 specification and the Java Websockets API. For the master version at qwait.csc.kth.se, Jetty version 9.1.1.v20140108 from the 8th of January 2014 was used. Versions of Jetty can be downloaded from here.

    Actually installing Jetty is dependent on your particular distribution and the version of Jetty chosen. One way to do it for Ubuntu is outlined in this blog post.

  3. If you want to use continuous deployment (so that the application can simply be deployed with mvn package cargo:redeploy from the source repository on any machine), download cargo-jetty-7-and-onwards-deployer from here under the "Tools" section, and put it in your Jetty webapps folder (e.g. /opt/jetty/webapps/). Configure the webapp according to the official instructions.

  4. Jetty listens to port 8080 by default. If the server should serve pages on port 80, it is recommended to either use a front-end proxy server such as Apache HTTPD or Nginx, or to introduce an iptables rule in Linux to forward port 80 to 8080, like so:

    /sbin/iptables -t nat -I PREROUTING -p tcp --dport 80 -j REDIRECT --to-port 8080
    

    Configuring this rule to be applied on each boot is distro-specific.

Configuration

The default QWait configuration "works" but is not optimal. You should change the settings of the application to suit your needs.

The web application is configured using the Spring settings API. This API allows settings to be overridden via different sources.

Before deployment

The easiest way to change a setting is to modify the file src/main/resources/settings.properties. This is a simple key=value settings file, with the additional property that Maven variables can be interpolated with ${...} syntax. A typical configuration file for a production environment, along with comments explaining each setting, would be:

# The name of the application to display in various places.  Defaults to
# <project><name>Whatever this is</name></project> from the pom.xml file.
product.name=${project.name}

# The version of the application to display in various places.  Defaults to
# <project><version>Whatever this is</version></project> from the pom.xml
# file.
product.version=${project.version}

# The driver to use when connecting to a database.  Here,
# PostgreSQL has been configured.  Note that a large number of
# databases are supported, but you need to make sure that the
# relevant JDBC driver has been added to the <dependencies> in
# pom.xml (For PostgreSQL, it is org.postgresql:postgresql)
dataSource.driverClassName=org.postgresql.Driver

# The URL to connect to, which is database specific.  In this
# case, we connect to PostgreSQL on localhost, with the database
# name qwait.
dataSource.url=jdbc:postgresql://localhost/qwait

# The user and password to use when connecting to the database.
dataSource.username=qwait
dataSource.password=<secret>

# Which SQL dialect to use when communicating with the database.
# Supported dialects are listed here: http://docs.jboss.org/hibernate/orm/4.0/javadocs/org/hibernate/dialect/package-summary.html
hibernate.dialect=org.hibernate.dialect.PostgreSQL9Dialect

# How to generate the schema for the database.  Supported values are:
# - validate: Don't do anything, just verify that the tables are
#   compatible with the webapp and error out otherwise.
# - update: Create new tables and columns if they are missing.
# - create: Overwrites existing tables with empty tables.
# - create-drop: Like create, but also drops all tables at application
#   shutdown.
hibernate.hbm2ddl.auto=update

# The CAS call-back service URI to use when doing CAS authentication
security.cas.service=http://qwait.csc.kth.se:8080/authenticate
# The CAS login URL to use
security.cas.loginUrl=https://login.kth.se/login
# The CAS logout URL to use
security.cas.logoutUrl=https://login.kth.se/logout
# The actual CAS ticket validator service
security.cas.ticketValidator=https://login.kth.se
# Our auth provider key for this webapp
security.cas.authProviderKey=<secret>

# The LDAP service to use when doing LDAP lookups
security.ldap.url=ldaps://ldap0.csc.kth.se
# The LDAP base query to constrain results
security.ldap.base=ou=unix,dc=csc,dc=kth,dc=se
# LDAP auth user if needed
security.ldap.userDn=
# LDAP auth password if needed
security.ldap.password=
# Enforce read-only operations via anonymous log-in
security.ldap.anonymousReadOnly=true

# Application configuration profiles to activate. If the "ldap"
# profile is activated, the application will try to use LDAP to
# look up user names.
spring.profiles.active=default,ldap

After deployment

When the application has been packaged as a .war, it is tedious to modify the settings inside of it (but possible: simply open the file as a ZIP archive).

Instead, settings can be overridden using container-specific mechanisms. For example, one way is to set Servlet context parameters. In Jetty, this is done globally by changing /opt/jetty/etc/webdefault.xml. A configuration setting might look like:

<context-param>
    <param-name>hibernate.dialect</param-name>
    <param-value>org.hibernate.dialect.PostgreSQL9Dialect</param-value>
</context-param>

It is also possible to do per-application overrides in Jetty using the override-web.xml file as outlined here.