This can be no trival task! All ids and dates are accurate as of 02/20/2018. This assumes you have access to the Bioconductor EC2 instances/snapshots and have a locally checkout version of the website code
It is suggested to make an EC2 testing environment where a test [git] branch can be created. Once that is debugged the testing [git] branch can be pulled locally to again test and debug if need be. Then it can be deployed.
Make a testing EC2 instance off the latest snapshot of staging.bioconductor.org.
The snapshot is taken every Friday.
1. Go to our Bioconductor EC2 instances: https://console.aws.amazon.com/ec2/v2/home?region=us-east-1#Instances:sort=desc:launchTime
2. Get the `staging.biconductor.org` instance id (i-6552ca8c)
3. Search for that instance id in the list of create Snapshots: https://console.aws.amazon.com/ec2/v2/home?region=us-east-1#Snapshots:sort=desc:startTime
4. Choose the most recent and launch
This launched EC2 instance is now our testing environment! This instance will have
the same account users and passwords as the original staging.bioconductor.org
This disables any tasks or interaction with the main EC2 instances. Login as each
user and disable the crontab immediately!! Then log on as biocadmin for all
updates.
Log in as biocadmin to your testing instance.
The staging environment uses rbenv to install and update ruby. See either https://github.com/Bioconductor/bioconductor.org#ruby or https://github.com/rbenv/rbenv
1. cd .rbenv
2. git pull
3. cd plugins/ruby-build
4. git pull
5. cd /home/biocadmin/bioc-test-web/bioconductor.org
6. rbenv install -l # this will list all available versions
7. rbenv install <ruby version> (ex: rbenv install 2.5.0)
8. rbenv global <ruby version> # this will set the default ruby version
I ran into ruby installation ERRORS such as the following:
rbenv: cannot rehash: /home/biocadmin/.rbenv/shims/.rbenv-shim exists
If this occurs simply delete this file AND the partially failed download of the version and try to download desired version again.
EX. If this occurred during the 2.5.0 installation, run the following
rm /home/biocadmin/.rbenv/shims/.rbenv-shim
rm -rf /home/biocadmin/.rbenv/versions/2.5.0
rbenv install 2.5.0
rbenv global 2.5.0
To install irb needed to test things interactively,
RBENV_VERSION=<ruby version> gem install irb
Updating gems at the very least will create an updated Gemfile.lock in the
directory /home/biocadmin/bioc-test-web/bioconductor.org. Website files or code
may also need to be updated with newer versions of required gems or debugged if
gems changed internal functionality or object definitions. It is recommended to start
a new branch early to keep track of all changes and merge once debugged.
git checkout -b updatingRubyTesting
On the testing EC2 instance, make sure you are using the version of ruby just installed. Once you remove the Gemfile.lock, there is no "version" attached to the gem, so ruby will just install the latest version of the gem and create a new Gemfile.lock.
NOTE: This will update all the gems.
ruby --version
rm Gemfile.lock
bundle install
Dependabot on github will periodically issue some security alerts, and suggest updating the gems. A good way to test the gems update is, to attach a version next to the gem in the Gemfile.
Eg: In the Gemfile
gem 'rack', '5.0.0'
Then,
bundle install
This will update the Gemfile.lock for you as well. Remember the Gemfile.lock is an autogenerated file, and should not be manipulated.
Another thing to keep in mind is, the previous gems which were installed and are not in use anymore should be removed, with the command,
bundle clean --force
This will remove the previously installed gems. This needs to be done whether updating a single gem or multiple gems.
You now should be able to rake with the updated version of ruby and gems. Running
rake in /home/biocadmin/bioc-test-web/bioconductor.org will start to build the
website locally - as our code is defined in a directory output
You can double check the version and installed gem list with the following:
ruby --version
gem list
NOTE There is a website script that is used that accesses support.bioconductor.org. This will fail unless the IP is added to the inbound rules of the EC2 instance for support.bioconductor. Since the Testing EC2 instance will have a different IP each time in is stopped and restarted it is recommended to disable this script in the Rake file for testing and debug locally if necessary. You may need your local IP added to the inbound rules; talk to whomever necessary that is managing these resources for this change.
# In Rakefile comment out the line at the top of the file
# require './scripts/get_post_tag_info.rb'
For good measure you can delete the output/ directory to make sure it is clean before doing the rake.
cd output
rm -rf *
cd ..
rake
Debug and update if necessary. If it is taking longer than expected (a few days, a week or two) you should try and update the gems to make sure all are still current by doing a gem update. NOTE You still MUST delete the Gemfile.lock if you want the gems to be updated
rm Gemfile.lock
gem update
Push your local git testing branch so that you can pull down all the changes locally and retest.
git push -u origin updatingRubyTesting
This also assumes you have a locally checked out version of the
website code found at:
https://github.com/Bioconductor/bioconductor.org Follow the same steps
above for updating ruby locally and git pull the testing branch that
was uploaded from the EC2 testing. You should pull down a new
Gemfile.lock that will have the updated gem versions. If bundler needs
to be updated, you can use bundle update --bundler which will also
install all your gems as well as update bundler.
cd bioconductor.org # your local checkout of website
git checkout -b updateRubyTesting origin/updateRubyTesting # pull branch
bundle install
You now should be able to rake with the updated version of ruby and gems. You can double check the version and installed gem list with the following as we did before:
ruby --version
gem list
For good measure you can delete the output/ directory to make sure it is clean before doing the rake.
cd output
rm -rf *
cd ..
rake
This may seem like overkill but feel like it is a good precautionary step. Once The testing branch is debugged locally, Merge the local testing branch into devel.
git checkout devel
git merge updatingRubyTesting
Make sure the Gemfile.lock was updated to the most recent versions of everything and make sure you can rake again without ERROR. Debug if necessary. If this goes as planned, push changes to github and git.bioconductor.org repository.
Note Be sure you are ready to deploy if the changes are pushed to the
git.bioconductor.org location. The repo will start being pulled immediately on
the next scheduled build of the website - This will mean there could be a period
of time where the website is broken if it pulls the changes but the gems and ruby
haven't been updated yet!
Log into the staging.bioconductor.org EC2 instance as biocadmin. Follow the
same instructions above for updating Ruby. Navigate to the website code
/home/biocadmin/bioc-test-web/bioconductor.org and do a git pull. Ensure that
the Gemfile.lock was updated to the newest versions and run the following:
gem install nanoc
gem install bundler
bundle install
You now should be able to rake with the updated version of ruby and gems. You can double check the version and installed gem list with the following as we did before:
ruby --version
gem list
To see if the updates are successful, instead of manually running rake, it is
recommended to tail the log of the website build and monitor there if it is
successful when staging does its scheduled tasks:
tail -f /home/biocadmin/bioc-test-web/log/update_site.log
Word of Caution Many Bioconductor systems are inter-related. After updating monitor that the website is building but also keep tabs on other systems that may connect to staging. A key example: When updating the sequel gem, the SPB was affected do to tighter regulations and when connections were tested.