-
Choose an Ubuntu-based instance, as Ubuntu readily comes with
apt-getand Python. We recommend researching Ubuntu's ideal version with its end of life plot. For example:v16.04is ideal for the year 2018. -
AWS Security groups need to have a TCP rule to open port number 8787.
-
Install Docker-CE onto the instance if it's not already available.
-
Allow for non-root users to manage Docker.
-
Install
docker composeby installing the Compose plugin:DOCKER_CONFIG=${DOCKER_CONFIG:-$HOME/.docker} mkdir -p $DOCKER_CONFIG/cli-plugins curl -SL https://github.com/docker/compose/releases/download/v2.18.1/docker-compose-linux-x86_64 \ -o $DOCKER_CONFIG/cli-plugins/docker-compose chmod +x $DOCKER_CONFIG/cli-plugins/docker-composeYou can test the installation with
docker compose version -
Clone the repository containing the miniDREAM course modules onto the instance. For example:
git clone --branch minidream2023 https://github.com/mc2-center/minidream-challenge.git
When building the server, it will expect the directory of course materials to be named
minidream-challenge. If you are not using minidream-challenge as the source of course materials, editdocker-compose.ymlso that the correct volume is mounted. See Shared Files for more details. -
Clone this repository onto the instance:
git clone https://github.com/mc2-center/minidream-r-env.git
-
Switch to the
minidream-r-envdirectory and start the server:cd minidream-r-env/ docker compose up --build -dIt will take around 15-20 minutes to build the server for the first time. Once complete, you should see the following:
... [+] Running 3/3 β Network minidream-r-env_default Created 0.0s β Container rstudio Started 1.6s β Container proxy Started 0.6s
Congrats! π The interactive RStudio environment is now available at http://minidream.synapse.org/.
From this point forward, you can:
-
start the server with:
docker compose up -d
-
stop the server with:
docker compose down
Note: stopping the server will reset the RStudio contents, e.g. modules will be removed, etc.
The RStudio server will come pre-installed with the following packages/libraries:
- BiocManager
- bitops
- caTools
- clusterProfiler
- DOSE
- getPass
- here
- imager
- org.Hs.eg.db
- pathview
- revealjs
- survival
- tidyverse
- tsne
To install a new package to the server, do:
docker compose exec -it rstudio R -e "install.packages('<package name>')"
By default, the server is configured to bind-mount minidream-challenge as
/shared in the rstudio container:
docker-compose.yml
...
volumes:
- ../minidream-challenge:/shared
If you list the files of /shared with docker compose exec -it rstudio ls -l /shared,
the files listed will match the minidream-challenge directory.
When a volume is mounted, it will preserve the same permissions it has on the
host. For example, let's say minidream-challenge has read-write-execute
enabled for everyone (user, group, other) on the host; when it gets mounted,
/shared will also have read-write-execute enabled for everyone.
To ensure that only the admins are allowed editing rights to the course materials on RStudio, change the group ownership as well as their permissions:
docker compose exec rstudio chgrp -R rstudio-admin /shared \
&& docker compose exec rstudio chmod g+rw /shared
Check the group ownership of /shared (just in case) with docker compose exec -it rstudio ls -l /shared
When the server is first built, two default users are added - one instructor (admin), one student:
admin,changeme,rstudio-user;rstudio-admin
student,changeme,rstudio-user
where:
- the first field is the username, e.g.
admin - the second field is the password, e.g.
changeme - the third field is the user's group(s), e.g.
rstudio-user;rstudio-admin
-
Assuming you are still in the
minidream-r-envdirectory, removerstudio/utils/users.csv(which is currently a symbolic link ofusers.csv.template):rm rstudio/utils/users.csv -
Create a new
users.csvwithin therstudio/utils/directory. Each line should contain 3 fields, delimited by a comma (no spaces!):- username
- password
- group(s)
If the user is a student, their group should be
rstudio-user. If the user is an admin, their groups should berstudio-user;rstudio-admin. See the default users list above for an example. -
Stop the server and rebuild:
docker compose down docker compose up --build -d
This time, the server should be ready in less than a minute (assuming no changes have been made to the server Dockerfile). Once ready, go to http://minidream.synapse.org/ and try:
- Logging in as a student
- Logging in as an admin
- Changing the password of a sample user with
passwd(in the terminal)
If the course is already in progress, you can add new users with the add_users
tool.
-
Create a new CSV file, following the same format as above.
-
Copy the CSV into the
rstudiocontainer:docker cp <new user list> rstudio:<new user list>You can check that the file has been copied over with:
docker compose exec -it rstudio ls -
Add the new users to the server:
docker compose exec rstudio /root/utils/add_users.sh <new user list>You should now be able to log in as one of the newly added users.
When the server is first built, none of the modules from the minidream-challenge
directory will be available on RStudio - each module will need to be shared
with the users. We recommend sharing one module at a time, following the same
pace as the miniDREAM course.
Let's go through an exercise of broadcasting a module to RStudio, starting with Module 0.
For a quick overview, list the available modules that can be broadcasted. For example:
$ docker compose exec -it rstudio ls -l /shared/modules
total 32
drwxrwxr-x 4 admin rstudio-admin 4096 May 25 23:09 module0
drwxrwxr-x 3 admin rstudio-admin 4096 May 23 17:07 module1
drwxrwxr-x 3 admin rstudio-admin 4096 May 23 17:07 module2
drwxrwxr-x 3 admin rstudio-admin 4096 May 23 17:07 module3
drwxrwxr-x 3 admin rstudio-admin 4096 May 23 17:07 module4
drwxrwxr-x 3 admin rstudio-admin 4096 May 23 17:07 module5
drwxrwxr-x 3 admin rstudio-admin 4096 May 23 17:07 module6
drwxrwxr-x 3 admin rstudio-admin 4096 May 23 17:07 welcome
-
Assuming you are in the
minidream-r-envdirectory, run thebroadcast_moduletool:docker compose exec rstudio root/utils/broadcast_module.sh \ shared/modules/module0 \ rstudio-userThis will share the contents of
shared/modules/module0to all users belonging to therstudio-usergroup (which should be all of the users in RStudio). Amodulesdirectory should now be available in the Files pane in RStudio. -
(One-time only) Some course materials are dependent on scripts and data located in
/home/shared/*, which does not exist when the server is first built. Create a symbolic link in therstudiocontainer so that this filepath exists:docker compose exec -it rstudio ln -s /shared /home/shared
Congrats! π You just launched your first module!
-
From the home directory, pull the changes:
git pull minidream-challengeAlternatively, changes can be directly applied inside the directory on the instance:
vim minidream-challenge/modules/<module name>/<notebook> -
Copy the new module contents to the container. For example:
docker cp \ minidream-challenge/modules/module0/intro-to-RStudio.Rmd \ fc9ac0f0f15f:/shared/modules/module0/. -
Switch to the
minidream-r-envdirectory and re-broadcast the module:docker compose exec rstudio root/utils/broadcast_module.sh \ shared/modules/module0 \ rstudio-user
Note: this section will assume you are using
minidream-challengeas the source for course materials.
-
In either the same instance or a new one, install the latest version of Miniconda:
mkdir -p ~/miniconda3 wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3 \ && rm -rf ~/miniconda3/miniconda.sh miniconda3/bin/conda init bash -
Exit the instance (to apply the changes), then log back in.
-
Assuming you are in the home directory, create a new environment from file:
conda env create -f minidream-challenge/environment.ymlThis will create a virtual environment called
minidreamthat will contain all of the necessary Python and R libraries to run the scoring harness. -
Activate the virtual environment.
conda activate minidream -
In the home directory, create a new file called
.synapseConfig, which will contain a Synapse Personal Access Token (PAT) needed to run the submission system:[authentication] authtoken=YOUR-PERSONAL-ACCESS-TOKEN-HERECredentials used should have editing rights to the miniDREAM challenge site. Go here to generate a new PAT.
Within minidream-challenge, edit the following scripts:
-
scoring_harness/challenge_eval.sh- Update the absolute path to the conda virtual environment on your machine
-
scoring_harness/challenge_config.py-
Update CHALLENGE_SYN_ID to the synID of the project for the challenge site.
-
Update CHALLENGE_NAME
-
Update the list of ADMIN_USER_IDS to contain the user IDs of
-
Update the evaluation IDs in
evaluation queues:evaluation_queues = [ { 'id': <need to be updated>, 'scoring_func': score, } ]
-
-
R/submission_helpers-
Update the folder synIDs where module submissions will be uploaded:
submission_folder <- switch( module, "0" = "syn000", "1" = "syn111", "2" = "syn222", "3" = "syn333", "4" = "syn444", "5" = "syn555", "6" = "syn666", "7" = "syn777" ) -
update the evaluation ID for the
synSubmitfunction:submission <- synSubmit(evaluation = "<need to be updated>", entity = activity_submission)
-
-
If it's not already active, activate the
minidreamvirtual environment. -
Assuming you are in the
minidream-challengedirectory, test whether you can runchallenge.pywithout any errors:python scoring_harness/challenge.py -h -
We also recommend testing out one of the commands as well, ideally, the same one listed in
challenge_eval.sh, e.g.python scoring_harness/challenge.py score --allIf it's working properly (and assuming there are no pending submission to be evaluated yet), you should get some logs in STDOUT like this:
=========================================================================== 2023-06-03T07:12:24.821643 Scoring: 9615336 - 2023 miniDREAM Module Submissions ------------------------------------------------------------ done: 2023-06-03T07:12:25.028427 ===========================================================================
Once you are sure the scoring harness is working as expected, set up a Cron job so that it will run every minute of every day.
-
Open up the crontab:
crontab -e -
Add the following task to the file:
* * * * * /home/<your username>/minidream-challenge/scoring_harness/challenge_eval.sh -
Save and exit the file. Assuming it's working properly, logs will be saved into
scoring_harness/log/score.log
Note: if errors are noted in the log file, we recommend stopping the cron job so that you can address the errors in the interim. You can temporarily stop the job by commenting out the task with #.
Congrats! π You just finished setting up the challenge infrastructure for miniDREAM!
-
rstudio-server --helpFor example:
rstudio-server suspend-allwill remove the message: "ERROR session hadabend" from an R session console after each service 'reboot'.
-
Check out leader board Check out the course page wiki. Check "scoreboards" section, and click on the relevant module. Then, click on "Wiki tools" widget -> "Edit Project Wiki". You should be able to uncomment the line related to leaderboard.
-
Clear submission After testing out your own submission, you might want to clear out previous submission. To do that, please use the
challengeutilspackage here
After installing challengeutils package, you could use challengeutils delete-submission <submission id> for deleting a certain submission.
- Update submission helpers
Check out file
submission_helpers.Runder folder R. This file by default could only edited by root user. You could edit it by following the steps:
- Interact with the docker container as a root user
docker exec -it rstudio bin/bash - Go to
/shared/Rfolder - Update
submission_helpders.Rthere by usingnanoNote: To save a file in nano, use^O(ctrlOon Mac)
-
Grant sudo access This could be done by using
sudo usermod -a -G sudo username -
Add users to docker group (Run docker commands without sudo) To be able to run docker without using
sudo, we will have to add users to the docker group. -
Check existing users in docker group
getent group docker
- Add a new user to docker group
sudo usermod -a -G docker username
Note: make sure that you are using jumpcloud username You should be able to see something like:
docker:x:999:ubuntu,username
- To activate the changes to the docker group without restarting docker daemon:
newgrp docker
Note: If the user is still getting error after the above steps, we might want to change group ownership of the /var/run/docker.sock file by using: sudo chown root:docker /var/run/docker.sock.
Reference could be found here and here
- Update forum link in
minidream-challenge/scoring_harness/messages.pyAfter students submit their work, they would get a message from the system. You could updatesupport_forum_urlandchallenge_instructions_url