Skip to content

Docker newbie instructions

Peter Staab edited this page Jun 21, 2019 · 54 revisions

-- by Mike Gage

Background

The docker files for webwork were constructed by Pan Luo (@xcompass) at UBC. Thanks Pan!!

This explanation is by and for newbies and may well be simplified, clarified, corrected in later editions. (Jan 6, 2018, July 21, 2018)

The purpose of using docker is to allow nearly everyone who has worked with webwork using the command line to download and test new patches or pull requests to WeBWorK. The initial setup takes a bit more than an hour. After that you can pull a new experimental version of WeBWorK in five minutes for review without interfering with any of the existing work on the computer. Faster testing will allow us to proceed more quickly with new WeBWorK features.

The ways of setting up Docker are evolving. What follows is a relatively simple (if slightly time consuming way) of installing WeBWorK inside docker. We are experimenting with installing webwork2, pg, the mysql database and OPL separately as docker components and then connecting them. This will make it easier to modify and save files in each of these components and to rebuild without having to rebuild everything. (Rebuilding the library database in particular takes some time.) Those with some experience using Docker and webwork may want to look at https://github.com/openwebwork/webwork2/wiki/Docker---things-to-get-used-to-(and-avoid-surprises) and review and contribute to experiments described in the develop branch and on the forum

Installing Docker


I have run docker on the MacOS high sierra and am now running it on Mojave. You need at least El Capitan MacOS in order to run Docker.

Docker CE has been successfully installed on Linux machines.

Docker CE is available for Windows machines as well, but getting WeBWorK running under Docker CE for Windows is a bit more complicated than doing so on MacOS X or Linux.

  • Docker CE for Windows depends on Hyper-V, so only works on Windows Professional, Education, or Enterprise, but not on Windows Home.
  • There is an older Docker Toolbox for Windows which works differently, and which may work on on machines whose version of Windows does not support Docker CE for Windows.
  • When Hyper-V is activated during the installation of Docker, it disables VirtualBox and other VM hypervisiors. (They cannot operate in parallel.)
  • In order to run VirtualBox again you have to disable Hyper-V. To do so: uncheck Hyper-V in Control Panel -> Programs -> Programs and Features -> Turn Windows features on or off. Then reboot.
  • Getting WeBWorK to work in Docker for Windows also requires:
    • getting symbolic links to work properly in Windows (possible on recent versions of Windows when Developer Mode is enabled)
    • and sharing the Windows drive on which you have your webwork2 folder.
  • See:

Please report on your experiences. This wiki is versioned so don't worry about making mistakes when editing this document -- we can always restore an older version.

All commands are issued from the command line so you will need a terminal window. You will need some experience typing unix commands in order to follow these instructions.

Install Docker

Install Docker on your computer. (There are versions for Mac, Windows and Linux)

Docker download home Get the community edition. (CE)

Docker download versions

The setup of docker on the Mac is standard and quick, see the Mac Instructions. For Windows see the Windows Instructions and https://servingmath.org/wp/2018/08/installing-webwork-on-docker-on-windows/

Testing the Docker installation

Test that it is installed by opening up a terminal and entering

$ docker --version
Docker version 18.06.0-ce, build 0ffa825

$ docker-compose --version
docker-compose version 1.22.0, build f46880f

$ docker-machine --version
docker-machine version 0.15.0, build b48dc28d

Do the "explore the application and run examples" section

mkdir  webwork_docker # (create a directory anywhere on your disk) 
cd webwork_docker # all your work will be in this directory
docker run -d -p 80:80 --name webserver nginx # start a local webserver

after a few minutes an image of the nginx image will have been pulled from the cloud and a container running the webserver will have been created from the image.

In a web browser type http://localhost/ and return (see the Mac Instructions, Windows Instructions, pages above for expected response). Something along the lines of "welcome to nginx".

docker ps    # see what containers are running, the first column is container ID
   gage-imac-461:webwork_docker mgage$ docker ps
   CONTAINER ID        IMAGE               COMMAND                  CREATED             STATUS               PORTS                NAMES
   017118b37fef        nginx               "nginx -g 'daemon of…"   2 minutes ago       Up 2 minutes        0.0.0.0:80->80/tcp   webserver

docker logs CONTAINER_ID # print log/stdout of the container with CONTAINER_ID
   gage-imac-461:webwork_docker mgage$ docker stop webserver
   webserver
docker stop webserver # stop the server

docker ps    # no container is running (they are all stopped  and therefore "hidden").

docker ps -a # includes all of the containers including ones that are stopped.
   gage-imac-461:webwork_docker mgage$ docker ps -a
   CONTAINER ID        IMAGE               COMMAND                  CREATED             STATUS                          PORTS               NAMES
   017118b37fef        nginx               "nginx -g 'daemon of…"   5 minutes ago       Exited (0) About a minute ago                 
   webserver
docker start webserver # restart the server

docker rm -f webserver # remove the webserver (the -f stands for "force" )

# the image which you downloaded from the cloud and from which the container was created is still on your machine

docker images
   gage-imac-461:webwork_docker mgage$ docker images
   REPOSITORY          TAG                 IMAGE ID            CREATED             SIZE
   nginx               latest              7042885a156a        44 hours ago        109MB

docker rmi nginx  # will remove the image completely

At this point you can read/skim the rest of the Mac Instructions, Windows Instructions, etc pages at your leisure. There are no changes that need to be made, but skimming to see what is available on this page can be helpful.

There is a second getting started page that is useful to read in order to get oriented. For example it explains the difference between a container and an image but first let's actually set up a webwork installation using docker.

Download and install a WeBWorK container:

Github and git activity

  • The two Git config settings included below -c core.autocrlf=false -c core.symlinks=true should really only be needed on Windows (as these are the default values on Linux and MacOS X), however forcing them should not interfere with anything on Linux and MacOS X.
  • It is not necessary to install the pg repository below. You can skip that git clone instruction unless you plan to make changes to the core pg code, in which case you will also need to edit webwork2/docker-compose.yml to include your local pg tree at run-time. ( See https://github.com/openwebwork/webwork2/wiki/Docker---things-to-get-used-to-(and-avoid-surprises))
cd webwork_docker
ls  # it should be empty -- running the demo above left no trace in this directory

git clone https://github.com/openwebwork/webwork2 -c core.autocrlf=false -c core.symlinks=true 
  # grab the webwork2 repo

#optional -- not needed until you start modifying pg
# git clone https://github.com/openwebwork/pg       -c core.autocrlf=false -c core.symlinks=true 
# grab and install the pg repo

cd webwork2
# Change to the branch needed. (Docker is now part of the master branch so no change is needed. 
# If you are doing advanced work you may want to switch to the develop branch.

git branch 
 * master               
# should show  "* master"  highlighted.
  • Depending on how fast and how accurately you type these git commands will take 5 minutes or less.
    • The revised files will run OPL-update automatically when the container is first run.
      • this takes 20 minutes to complete.

Docker commands

Before proceeding, shut everything down in Docker and clear images. docker ps -a should be clear.

We are still in the webwork2 directory:

ls

If everything is successful the ls listing will show the usual webwork files but also the files
Dockerfile, docker-compose.yml and docker-entrypoint.sh

Now while still in the webwork2 directory type

docker build .
docker-compose up 

You will now have to wait for a while. The terminal window will scroll through lots of lines to let you know that something is running.

Next time you will probably prefer to run

docker-compose up -d   # detaches the terminal so you don't see all the output
                     # and so that you can quit the terminal without stopping the container.

You can see that it is installing a lot of things for the ubuntu version of the linux operating system. Eventually it installs TeX and Apache, various CPAN modules (you'll see a lot of Building and testing lines) and then MathJax. Then there are lines about creating the mysql (dabase) application db_1 and finally the webwork application app_1. Then the admin course is created and then it starts creating the OPL database.

IMPORTANT: Don't close the terminal window since that will stop the container running the website.

Towards the end you should see

........
app_1  | Admin course is created.
app_1  | create modelCourse subdirectory
app_1  | defaultClasslist.lst is being created
app_1  | Library version is 2.5; using OPLtables!
app_1  | Mysql database reinitialized.
app_1  | Reading in textbook data from Textbooks in the library /opt/webwork/libraries/webwork-open-problem-library/OpenProblemLibrary.
app_1  | Reading in OPL taxonomy from Taxonomy2 in the library /opt/webwork/libraries/webwork-open-problem-library/OpenProblemLibrary.
app_1  | Saved taxonomy to /opt/webwork/webwork2/htdocs/DATA/tagging-taxonomy.json.
app_1  | Converting data from tagged pgfiles into mysql.
app_1  | Number of files processed:
app_1  |    100   200   300   400   500   600   700   800   900  1000
.....

as docker-compose completes the installation of webwork (app_1), the database (db_1) and R (r_1) and it starts to build the OPL database. Creating the OPL database takes the most time. There are about 35K files that have to be processed. Go get some coffee. After the files are processed there are 126 textbooks that need to be processed to line up problems relevant to those text books.

On my iMac the whole process took about 35 minutes from typing docker build . ; docker-compose up; to the final result

app_1  | Updating Library Statistics.
app_1  | Importing statistics for 2 courses.
app_1  |    1   2
app_1  |
app_1  | mysql: [Warning] Using a password on the command line interface can be insecure.
app_1  |
app_1  | Done.
app_1  | Fixing ownership and permissions (just in case it is needed)
app_1  | AH00558: apache2: Could not reliably determine the server's fully qualified domain name, using 172.18.0.4. Set the 'ServerName' directive globally to suppress this message
app_1  | webwork.apache2-config:  WeBWorK server is starting
app_1  | webwork.apache2-config:  WeBWorK root directory set to /opt/webwork/webwork2 in webwork2/conf/webwork.apache2-config
app_1  | webwork.apache2-config:  The following locations and urls are set in webwork2/conf/site.conf
app_1  | webwork.apache2-config:  PG root directory set to /opt/webwork/pg
app_1  | webwork.apache2-config:  WeBWorK server userID is www-data
app_1  | webwork.apache2-config:  WeBWorK server groupID is root
app_1  | webwork.apache2-config:  The webwork url on this site is http://localhost/webwork2
app_1  | webwork.apache2-config:  The webwork smtp server address is localhost
app_1  | webwork.apache2-config:     The webwork smtp server port is
app_1  | webwork.apache2-config:     The webwork smtp server protocol is 'not ssl'
app_1  | WebworkSOAP::WSDL: webwork_directory set to /opt/webwork/webwork2 via $WeBWorK::Constants::WEBWORK_DIRECTORY set in webwork.apache2-config
app_1  | WebworkSOAP::WSDL: rpc_url set to http://localhost/webwork2_rpc
app_1  | WebworkWebservice: webwork_directory set to /opt/webwork/webwork2 via $WeBWorK::Constants::WEBWORK_DIRECTORY set in webwork.apache2-config

DO NOT CLOSE THIS WINDOW!!!! The window must be kept open while running the webserver. If you close the window it will shut down the webwork server.

Testing out the locally running WeBWorK container

Once the lines have stopped scrolling you can type http://localhost:8080/webwork2 into your browser and you have a webwork site running.

The only course is "course administration" and it has login "admin" with password "admin".

You can login and add yourself as another admin in the admin course. (First add yourself as a student, then promote yourself to admin.)

You can check the library browser to make sure that it works and that the OPL database has been updated properly.

Create a new course 'test_course', add some guest "students" by importing from demoCourse.lst and add a few homework problems to the course using the LibraryBrowser.

What could possibly go wrong?

  • I forgot to select a target set from the course in the library browser. Usually the blank for the target set would read "Select a Set from this Course" but because no homework sets had been set up at all this field was set to "no sets in this course yet". This lead to an "instructorXMLRPC: 403 Forbidden" error.

    • Usually "instructorXMLRPC: 403 Forbidden" error is triggered by some glitch in setting up the site URL in the site.conf file so that it works properly with AJAX, but not in this case. Making sure that I had created a new set in the course and then choosing that set as a target was enough to clear the error.

    • I found a message in the error log at /var/log/apache2/error.log indicating that no homework set had been created for the course (see more information below on reading this log) but that message was not transmitted to the screen. It's likely that this particular error can only occur with a brand new, empty course.

  • On some builds I got an "app_1 exited with error 1" at the end of the "docker build ." command or the "docker-compose up". This is obviously not good. In some cases I had to go back to the beginning and start the reinstall from the beginning. This involved removing all prebuilt docker packages so docker container ps -a and docker images should all be empty. If they are not use docker container rm <id> and docker rmi <id> to remove them. You only need to type the first couple of numbers/letters of the SHA identifier -- not the whole thing.

    • The docker set up is still fragile. We hope to break it into pieces so that webwork2, pg, mysql and OPL are all handled separately. This way if one download fails you don't have to start the entire process over. In particular you can update changes to webwork2 and pg without necessarily rebuilding the OPL database.

Next step

In a new terminal window go to webwork_docker/webwork2. Don't close the first terminal window.

Where is the data being stored? You won't be able to see much of the data that you usually expect to see stored in webwork on the disk, at least not from "outside" the docker environment. One thing you can see is the courses directory.

A new directory has been created to hold persistent webwork2 data

cd webwork_docker
ls ww-webwork-data
   courses
ls ww-docker-data/courses
  admin/			adminClasslist.lst	defaultClasslist.lst	modelCourse/		test_course/

You'll see courses, which is where persistent data for courses run in the Docker container will be saved.

IMPORTANT: You can add (upload) archived courses to the courses directory and then unarchive them using the admin course in the browser. This is the quickest way to populate your docker webwork with courses you have been using on your production site. Even if you don't have command line access to your production site you can archive and download your own courses using the commands on the FileManager page of your course.

With older versions of the Docker config files, there will also be a db directory which is the location for the persistent storage for the database. More recent versions of the Docker config files put the databases into a named Docker volume. It's wise not to touch the db directory unless you are a MariaDB expert.

Read more about the persistent and transient storage at https://github.com/openwebwork/webwork2/wiki/Docker---things-to-get-used-to-(and-avoid-surprises). It's one of the things that takes a while to get used to. The exact balance of which data is persistent and which is transient will continue to be tweaked as we work on creating the best experience of using docker with WeBWorK.

In your second terminal window (in directory webwork_docker/webwork2) -- you may have to cd to webwork_docker/webwork2 -- type

docker-compose down
   Stopping webwork2_app_1 ... done
   Stopping webwork2_db_1  ... done
   Stopping webwork2_r_1   ... done
   Removing webwork2_app_1 ... done
   Removing webwork2_db_1  ... done
  Removing webwork2_r_1   ... done

If you go back to your first window you will see that the build process has exited and your browser can no longer access the webwork site.

congratulations you've completed your first use of webwork on your laptop using docker.

Next try -- things will be quicker now:

In the first window type

docker-compose up -d
    Creating network webwork2_default with the default driver
    Creating webwork2_db_1 ... done
    Creating webwork2_r_1  ... done
    Creating webwork2_app_1 ... done

Notice that things are much quicker this time as the images are already built. And the webserver has started up right where you left off.

Close the first terminal window. And check that the website is still running. (This is what the -d accomplished.)

In the second terminal (webwork_docker/webwork2 directory) type

docker container ls  
# (or docker ps will also work) 

Observe the container id for the webwork container. Some commands will require at least the first 4 or 5 characters of that WeBWorK_id to identify the container. Sometimes the name webwork2_app_1 alone will work. You will need this in the command below.

type

docker container exec -it WeBWorK_id bash
     root@WeBWorK_id:/opt/webwork#

A terminal window has just been opened to the inside of the "remote" machine at the directory /opt/webwork/webwork2.

This command is worth creating an alias for. In my .bashrc account I have added

alias dockerbash="docker container exec -it webwork2_app_1 bash"

so that I don't have to remember the whole string. This command lets you "login via a terminal window" to the webwork2 site running in the docker container. What you see should be similar to what you see when you login via a command line to your production server or a development server. One difference is that you also have an "outside" view via the desktop of the permanent files. Changes made on the outside, e.g. copying a course from your desktop into the ww-docker-data/courses folder can be seen (after a brief pause) via the terminal at /opt/webwork/courses

Anything that you would do by logging in to your local WeBWorK server on the command line you can do by "logging" in to the container using the approach above.

Warning: Recall that changes made in the command line of the Docker container will only be persistent if they are making changes to files on the persistent storage. See https://github.com/openwebwork/webwork2/wiki/Docker---things-to-get-used-to-(and-avoid-surprises). Courses are on permanent storage, changes to macro files in pg or in the OPL are not. Changes to a macro file in a course template/macros file will be preserved.

When you are done exploring type exit and you will close the terminal window "connection" to the container (the remote machine)

Installing an archived course:

If you have a copy of an archived course that you've downloaded from your usual webwork site (a .gz file) we can place that in the container as follows.

  1. In webwork_docker/webwork2
docker-compose down
cp ../path/to/course/archiveFile.tgz ww-docker-data/courses   
docker-compose up
  1. Check the web browser and the admin course. You should be able to "unarchive" your course using the "unarchive" tab.
  2. If it is an old course you may have to "update courses" to fix the database on the course you have imported.
  3. In the web browser open the library browser and make sure that it works.
  4. You can now add more problems to that course. Changes will be preserved.

Misc:

Double check that you are running the versions of webwork and pg that you want. At the bash prompt (which you opened inside the Docker container) type:

cd pg
git fetch origin   # will refresh you copy of the index 
                   # of the branches available at origin (aka github.com/openwebwork)
git branch -a          
git checkout -b develop origin/develop  # if you want to test the develop branch
apachectl graceful # restarts the web server.

You do not need to rebuild the entire docker container to change from one branch of of webwork to another. (whew!)

Adding a new course

  • Now try adding a new course from the browser using the Course Administration web page.
  • You should see that you can build it using modelCourse as a template.

Shutting down (and restarting) the Docker container:

  • When you are done using your Docker container type docker-compose down and docker will stop using CPU's on your computer.
  • It will still take up some disk memory space.
  • It can be restarted with docker-compose up -d .

Final comments:

  • see Docker for newbies part 2 for adding daemon_course (still under development)

That's enough for now. Play with the WeBWorK site you have created and go back to skimming Docker's getting started docs to try to figure out what you have just done. We've only used the first two sections of that document -- there is more that you can do to create a "production site" that can handle many students.

Warning With what we have so far you DO NOT want to use this as a production WeBWorK site -- this is a local installation on your machine for your use only. None of the databases are using passwords for example, so there would not be much security if you start letting people do homework on a network connected to your computer.

To repeat, the purpose of using Docker (so far) is to allow nearly everyone who has worked with WeBWorK using the command line to download and test new patches or pull requests to WeBWorK. Faster testing will allow us to proceed more quickly with new WeBWorK features. You can for example create homework questions that use the latest features in the develop branch without endangering the behavior of your school's production WeBWorK site.

The UBC people are eventually planning to use docker on their production machines so that it is super simple to install WeBWorK.

Also watch out, this installation is temporary. If you create a WeBWorK course in your local site it will stay around if you do docker-compose down, which just makes the container dormant. It will also stay around if you quit the Docker app. If you remove the container however (a command with a "rm" or "rmi" in it) the site will disappear. You could first save the course using the "archive" button in the FileManager in WeBWorK and download the archived file. (You could also do this for multiple courses from Course Administration course.)

Cleaning up

All the persistent files (database and course files) are stored in webwork_docker/ww-docker-data directory. In order to do a completely clean uninstall of this "webwork in a box", remove the ww-docker-data directory or webwork_docker directory (and the named mysql storage volume).

You can’t perform that action at this time.