geoffreyvanwyk / learningsandbox.online Goto Github PK

This is necessary to:

• Prevent malicious users from using plugins to take over the server.
• Prevent the server's storage space being used up.

This can be done by setting $CFG->disableupdateautodeploy = true; in config.php.

The Ansible role for Moodle must be updated before this can be done.

This issue occurred when first installing Moodle 3.9 during deployment from GitHub Actions.

TASK [Copy fresh database 3.9] *************************************************
An exception occurred during task execution. To see the full traceback, use -vvv. The error was: DETAIL:  There is 1 other session using the database.
fatal: [***]: FAILED! => {"changed": false, "msg": "Database query failed: source database \"moodle-39\" is being accessed by other users\nDETAIL:  There is 1 other session using the database.\n"}

See https://github.com/geoffreyvanwyk/learningsandbox.online/actions/runs/6953214981/job/18918114038#step:7:241 .

The database cannot be copied while it is being accessed. This is probably due to the cron, which means the site has to be placed in maintenance mode before preparing for reset and disabled again afterwards. This also applies to the moodle_reset playbook.

This is much easier than for the user having to go back to the website to copy the credentials.

This will involve extending the Boost theme.

This is Moodle 4.3.

If I was on Twitter, I could just have linked to that.

I could add a link to LinkedIn. Most people are on there.

Otherwise, find if there is a way for having a Contact page on a static website.

The icons must be in the fonts directory according to the Line Awesome file structure.

The automatic reset needs to take place every hour. Originally the intention was to do the reset on the hour. The reset will be performed from GitHub Actions. GitHub Actions may delay or drop a job at high load times, such as on the hour. (See the note for the schedule event: https://docs.github.com/en/actions/using-workflows/events-that-trigger-workflows#schedule). For this reason, the reset should take place 30 minutes past the hour, every hour.

To make the reset instant, a fresh moodledata directory and a fresh database needs to be available at all times for each instance. The reset is then performed by a cron script triggering an Ansible playbook that updates config.php to point to the new directory and database. Thereafter, a new moodledata directory and database needs to be created again for the next switch. The old ones need to be deleted.

Update to Moodle playbook

• After installation of a Moodle instance, copy the moodledata directory.
• After installation of a Moodle instance, copy the database.
• Create GitHub Actions workflow for running Reset Playbook every hour.

Tasks for Reset playbook

• Rename current moodledata to old.
• Rename fresh moodledata to current.
• Rename "current" database to old.
• Rename next database to "current.
• Remove old moodledata.
• Drop the old database.
• Extract fresh moodledata.
• Copy next database.

This is also the case on Moodle HQ's demo site.

The reset objects are the extra moodledata directories and extra database tables.

The new domain has already been obtained.

The old domain should be redirected to the new domain.

Make it simple, e.g. just the letters LSO in a square or cirlce.

Use a hero element that states the usefulness of the site as in the README.

See https://umami.is/

This is to prevent malicious users using the Moodle instances to send spam.

This can be done in config.php by setting $CFG->noemailever = true;.

The Ansible role for Moodle needs to be updated to be able to set this with a variable.

Building the website with Gulp will be much easier than with HydePHP, because:

• with HydePHP, there are two systems: the PHP system and the NPM system. With Gulp, there is only Gulp.
• the php hyde serve clashes with npm run watch almost every time because they write to the same directory. With every second change almost, the npm run watch command crashes while trying to chmod the mix-manifest.json file. This is very bad for productivity. With Gulp, I will only run one command gulp watch.
• Autoreloading the web page in the web browser is much simpler with Gulp and BrowserSync. With HydePHP I would probably only figure it out after a lot of work.
• I only need one page, and maybe a "Contact" page. Gulp feels much lighter for this than HydePHP.
• HydePHP brings in styling customizations which interferes with DaisyUI.

Some plugins are almost always necessary on any Moodle installation, e.g. the Custom Certificate plugin. Because users cannot install plugins, these plugins should be installed.

• Make list of useful plugins.

If during a deploy of new commmits, new Moodle code is pulled from Moodle's GitHub, then the upgrade pages will appear after the reset has happened, and this will happen again after every reset. This is due to the fact that the saved database will not be upgraded.

Therefore, for each Moodle, the variable moodle_deploy_update must be set to no. Because the reset backups have already been made, all the installations must be redone.

• Set the moodle_deploy_update variable to no for each Moodle instance (12 minutes)
• Disable Reset workflow in GitHub Actions (1 minute)
• Reset Digital Ocean droplet to the ubuntu2204-nopasswd_sudo_user_ubuntu snapshot (1 minute)
• Run the Deploy workflow in GitHub Actions (50 minutes)
• Enable Reset workflow in GitHub Actions (1 minute)

The Reset workflow in GitHub Actions repeatedly failed. I have manually disabled it.

The setup for resetting Moodle 3.9 is now in disarray with regards to the moodledata and the databases.

This is due to stopping the Reset workflow after it started and not allowing it to complete. I had to do that while an automatic deployment was being initiated.

The manual fix for this in the future is to not merge pull requests around the time of the Reset workflow (which is 30 minutes past the hour).

Maybe there is a GitHub Actions way of solving this automatically.

The plan to fix this now is to just rebuild the server from the saved image on Digital Ocean.

This will allow users to see what emails look like.

For example, use: https://github.com/ryangjchandler/alpine-clipboard

See https://umami.is/

This should also include removing the default virtualhost because it can cause HTTP to not be redirected to HTTPS.

The README must mention why the Learning Sandbox is useful.

Free, instant access to default instances of Moodle versions for potential users who want to try it out, for existing users who want to see how Moodle behaves by default compared to their installations, for someone who quickly ones to try out a feature, and so forth.

The README must also mention the necessary limitations.

• removal of the ability to install plugins. The removal of that ability is to protect the server from malicious users who might use plugins to take over the server or completely use up all disk space.

• removal of the ability to send email so that malicious users cannot use the server to send spam.

• automatic reset to default of every instance on the hour, every hour.

These are 3.9, 3.11, 4.0. A value added service can be provided where private sandboxes are created for older versions.

Basically, a schedule should be added to the Deploy workflow for running once a week.

The workflow should run on the hour and disable the Reset workflow before starting, then reenable it afterwards.

This can be done by using a WSL instance.

Not sure how an image could get a permanent discoloring. Probably caused by the theme.

Part of this is to use a light theme. The theme does not have to match Moodle, which was what I was going for with the current theme.

This is the link on the right-hand side of the navigation bar represented by the GitHub icon.

The Notifications page could possibly be overridden using $CFG->customscripts.

This would mean rebuilding the site in Laravel.

A script can be added to the footer via Appearance > Additional HTML and it can be set via config.php to prevent users from removing it.

The variable $CFG->preventexecpath in config.php should be set to true.

I have fixed this on the server manually, but it must still be added to the playbooks.

The solution is to disable the default port 80 virtualhost: 000-default.conf.

See https://github.com/geoffreyvanwyk/learningsandbox.online/actions/runs/7741438795/job/21108475576#step:8:164

Instead of pulling from Twitter.

If the page is not reset, the timer will roll over and indicate that the reset in within an hour again, misinforming the user. The page should be refreshed via JavaScript. This will take the authenticated user back to the login page.