- [Improvement] Upgrade all services to open-release/lilac.2.
- [Bugfix] Fix "`sh` is not a recognized command" in some plugins, including minio.
- [Improvement] Set the default contact mailing email address
- [Bugfix] Fix minio initialisation in Kubernetes.
- [Bugfix] Fix "Invalid command argument" during upgrade from Koa to Lilac.
- [Bugfix] Fix mysql initialisation in docker-compose==2.0.0beta4.
- [Improvement] Tutor is now published on pypi as "tutor".
An issue with the latest release of docker-compose was reported here:
https://discuss.overhang.io/t/undefined-entrypoint-throws-error-in-docker-compose-2-0-0-beta-4/1716
The mysql-job definition had an empty entrypoint (`[]`). This was causing the following error:
the initiation of mysql fails with “services.mysql-job.entrypoint must be a string …
Error: Command failed with status 15”
I can't remember at all why we had to define an empty entrypoint. It probably
has to do with the fact that we could not run `sh -e -c "..."` commands in
mysql jobs. Similarly, the k8s job definition sets `command: []`. I tested both
local and k8s deployments without these definitions and they work just fine. So
I guess we can get rid of them.
The package maintainer of the "tutor" package was kind enough to
transfer ownership of the project to us. This is great, because we no
longer have to use the "openedx" suffix, which is trademarked.
For the time being, we keep maintaining the "tutor-openedx" package
which has a 1-to-1 dependency on the "tutor" package. In the future, we
expect that we will no longer push upgrades to tutor-openedx.
- [Bugfix] Fix double pulling mongodb image when upgrading from Koa to Lilac.
- [Improvement] Better logging during `plugins disable`.
- [Bugfix] Fix "upstream sent too big header" error during login of existing users after a Koa to Lilac upgrade.
- [Feature] Added the ability to skip `config.yml` file modification while running `tutor config save` command with `-e` or `--env-only` flag.
- [Feature] Add new config value `FORUM_MONGODB_DATABASE` to set the forum database name
In config.yml the new value FORUM_MONGO_DB_DATABASE was added with `cs_comments_service` as default value.
In docker-entrypoint.sh of forum I changed the hardcoded `cs_commecnts_service` with the new config value.
Multiple .yml files changed to handle the new config value.
When disable a plugin that set config entried, such as the minio plugin, tutor was logging the following:
Disabling plugin minio...
Removed config entry OPENEDX_AWS_ACCESS_KEY=openedx
Removed config entry OPENEDX_AWS_SECRET_ACCESS_KEY={{ MINIO_AWS_SECRET_ACCESS_KEY }}
Plugin disabled
The config values were not rendered during printing, which is a shame, because
the whole point of this log line is to warn users of passwords/secrets that are
being removed. Here, we make sure that the config values are properly rendered.
The new logs are now:
Disabling plugin minio...
Removing config entry OPENEDX_AWS_ACCESS_KEY=openedx
Removing config entry OPENEDX_AWS_SECRET_ACCESS_KEY=64vpCVLxhDxBuNjakSrX4CQg
Plugin disabled
- [Improvement] Avoid permission issues in Kubernetes/Openshift for users who do not have the rights to edit their namespace.
- [Improvement] Better Kubernetes object creation.
In most cases, it makes very little sense to edit the namespace that an
application is running in. Quite often, users are granted access to just one
namespace and don't have the necessary rights to edit the namespace -- and for
good security reasons. In such cases, the k8s namespace object already exists
and there is no need for the user to edit or create it. Here, what we do is
that we create the namespace only if it does not exist. This should solve quite
a few permission issues, notably for Openshift users.
- 💥[Security] Disable python-evaluated input by default as we don't run codejail.
- [Bugfix] Fix missing discussion notifications.
- [Improvement] Better error logging when loading a plugin from an incompatible version.
As described in issue #284, tutor does not come with codejail enabled out of
the box. Actually, we don't even have a working plugin, yet. To prevent users
from running unsafe code, we explicitely disable python-evaluated input by
disabling the "python" interpreter. This might break some courses; thus, this
is a non-backward compatible change.
Discussion email notification relies on edx_ace and celery routing
key for ace is not set properly by default, which causes the send
mail task is never consumed.
Close #439.
When upgrading Tutor plugins to the next release, I often end up with a
virtualenv that contains plugins that depend on different versions of
tutor-openedx. This causes a crash that did not log the name of the responsible
package. For instance:
Traceback (most recent call last):
File "/home/regis/venvs/tutor/bin/tutor", line 11, in <module>
load_entry_point('tutor-openedx', 'console_scripts', 'tutor')()
File "/home/regis/projets/overhang/repos/overhang/tutor/tutor/commands/cli.py", line 37, in main
add_plugin_commands(cli)
File "/home/regis/projets/overhang/repos/overhang/tutor/tutor/commands/plugins.py", line 137, in add_plugin_commands
for plugin in plugins.iter_installed():
File "/home/regis/projets/overhang/repos/overhang/tutor/tutor/plugins.py", line 401, in iter_installed
yield from Plugins.iter_installed()
File "/home/regis/projets/overhang/repos/overhang/tutor/tutor/plugins.py", line 362, in iter_installed
for plugin in PluginClass.iter_installed():
File "/home/regis/projets/overhang/repos/overhang/tutor/tutor/plugins.py", line 199, in iter_installed
for plugin in cls.iter_load():
File "/home/regis/projets/overhang/repos/overhang/tutor/tutor/plugins.py", line 233, in iter_load
yield cls(entrypoint)
File "/home/regis/projets/overhang/repos/overhang/tutor/tutor/plugins.py", line 220, in __init__
super().__init__(entrypoint.name, entrypoint.load())
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/pkg_resources/__init__.py", line 2442, in load
self.require(*args, **kwargs)
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/pkg_resources/__init__.py", line 2465, in require
items = working_set.resolve(reqs, env, installer, extras=self.extras)
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/pkg_resources/__init__.py", line 791, in resolve
raise VersionConflict(dist, req).with_context(dependent_req)
pkg_resources.VersionConflict: (tutor-openedx 11.2.11 (/home/regis/projets/overhang/repos/overhang/tutor), Requirement.parse('tutor-openedx<13.0.0,>=12.0.0'))
In this commit, we introduce an error log that displays the name and location
of the package. E.g:
Failed to load entrypoint 'minio = tutorminio.plugin' from distribution tutor-minio 12.0.0
- [Security] Apply Django security patches by upgrading from 2.2.20 to 2.2.23.
- [Bugfix] Fix video unit completion (see [pull
request](https://github.com/edx/edx-platform/pull/27230)).
- [Bugfix] Fix crashing installation because of a major release of all
Pallets projects.
- [Bugfix] Fix crash in `local quickstart -p` command.
- [Bugfix] Fix 502 error on request to lms with header larger than the
maximum uwsgi buffer size.
All pallets project requirement had a major upgrade today:
https://palletsprojects.com/blog/flask-2-0-released/
We are not yet compatible with click 8.0 and others. In particular,
click-repl imports modules which are no longer available. Until we can
upgrade, we add constraints to the requirements files.
The following error was being raised:
$ tutor plugins list
Traceback (most recent call last):
File "/home/data/regis/tmp/testtutor/bin/tutor", line 5, in <module>
from tutor.commands.cli import main
File "/home/data/regis/tmp/testtutor/lib/python3.6/site-packages/tutor/commands/cli.py", line 6, in <module>
import click_repl
File "/home/data/regis/tmp/testtutor/lib/python3.6/site-packages/click_repl/__init__.py", line 6, in <module>
import click._bashcomplete
ModuleNotFoundError: No module named 'click._bashcomplete'
Close #430.
When running `tutor local quickstart -p` we were getting the following error:
Usage: custom [OPTIONS] ARGS...
Try 'custom --help' for help.
Error: Missing argument 'ARGS...'.
The docker-compose command sometimes accept a single command ("pull") with zero
argument.
See: https://discuss.overhang.io/t/local-quickstart-not-working-when-pullimages-enabled/1526
According to [1], request to uwsgi with header
larger than the default value 4096 will end up
with below error as #426:
invalid request block size: 4123 (max 4096)...skip
hr_instance_read(): Connection reset by peer
[plugins/http/http.c line 647]
This commit fixes it by changing the maximum
buffer size as 4096
Also: Similar issue was already identifed and
fixed in [2] on tutor-ecommerce
[1] https://uwsgi-docs.readthedocs.io/en/latest/ThingsToKnow.html
[2] 6df2c99362
Close #426
During Docker images build process, apply custom edx-platform patches when tutor patch 'openedx-dockerfile-git-patches-default' is defined or apply current release patches in other case. It avoids possible conflicts between the actually used edx-platform version and the current release patches.
- [Improvement] Upgrade all services to open-release/koa.3.
- [Feature] Make it possible to build the openedx Docker image with a custom openedx-i18n version with the ``--build-arg OPENEDX_I18N_VERSION=custom`` command line argument.
We remove security patches and custom fixes which are now part of koa.3.
We take the opportunity to make it possible to build the openedx Docker image
without relying on a corresponding openedx-i18n repo tag: often, we want to
test whether the image simply builds successfully, and we don't need up-to-date
translations. For those cases, it's now possible to pass the `-a
OPENEDX_I18N_VERSION=oldertag` build argument.
When the PLUGINS config entry is None (`PLUGINS:`), the following error
was being triggered:
File "/.../tutor/tutor/plugins.py",
line 304, in is_enabled
return name in config.get(CONFIG_KEY, [])
TypeError: argument of type 'NoneType' is not iterable
- [Bugfix] Fix "Invalid Request" error during SAML authentication
(thanks @Ant1x!).
- [Feature] Add `images build --pull` option to update base images.
- [Improvement] Annotate types all over the Tutor code base.
- [Bugfix] Fix parsing of YAML CLI arguments that include equal "="
signs.
- [Bugfix] Fix minor edge case in `long_to_base64` utility function.
- [Improvement] Add openedx patches to add settings during build
process.
This option is mostly useless to us in CI, as it attempts (and fails) to
update the base image when building the dev image. For good reason: the
base image of the dev image is the latest openedx, which has not been
pushed to docker.io at the time of building -_-
Annotations were generated with pyannotate:
https://github.com/dropbox/pyannotate
We are running in strict mode, which is awesome!
This affects a large part of the code base, which might be an issue for
people running a fork of Tutor. Nonetheless, the behavior should not be
affected. If anything, this process has helped find and resolve a few
type-related bugs. Thus, this is not considered as a breaking change.
lms-worker was configured to run CMS tasks instead of LMS tasks. I'm not
sure what tasks were being dismissed, and what is the actual production
impact.
Redis data was not actually persisted, because the redis configuration file was
not mounted from the right location. In order to mount redis data in a
host-mounted directory, the working directory has to be properly set.
The problem was occurring both with docker-compose and Kubernetes.
Close #404.
When tutor is run as root, the `tutor config printroot` and `tutor
config printvalue` commands used to write to stdout, which was causing
issues when combined with `tutor config render` and `tutor local
settheme` commands.
See discussion:
https://discuss.overhang.io/t/error-while-install-indigo-theme/1056
First, allow using custom Django settings on a development
environment (as documented but not implemented), setting it to the
correct value of `tutor.development`. Prior to this, `tutor dev
runserver lms` would default to `tutor.production` when on a custom edX
branch.
Second, fix the documentation so the correct environment variable is
described, at the same time removing an option that doesn't seem to work.
See discussion: https://discuss.overhang.io/t/koa-dev-lms-doesnt-find-static-content/1250
We manage to get unit tests to run in a dedicated openedx-test container. Only
35 tests are failing (out of 17k). I suspect these tests are also failing in
the devstack.
`local exec` was failing with the following stacktrace:
Traceback (most recent call last):
File "/home/regis/venvs/tutor/bin/tutor", line 11, in <module>
load_entry_point('tutor-openedx', 'console_scripts', 'tutor')()
File "/home/regis/projets/overhang/repos/overhang/tutor/tutor/commands/cli.py", line 38, in main
cli() # pylint: disable=no-value-for-parameter
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/click/core.py", line 829, in __call__
return self.main(*args, **kwargs)
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/click/core.py", line 782, in main
rv = self.invoke(ctx)
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/click/core.py", line 1259, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx))
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/click/core.py", line 1259, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx))
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/click/core.py", line 1066, in invoke
return ctx.invoke(self.callback, **ctx.params) File "/home/regis/venvs/tutor/lib/python3.8/site-packages/click/core.py", line 610, in invoke
return callback(*args, **kwargs) File "/home/regis/projets/overhang/repos/overhang/tutor/tutor/commands/compose.py", line 237, in ex
ecute dc_command.callback("exec", args)
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/click/decorators.py", line 33, in new_func
return f(get_current_context().obj, *args, **kwargs)
File "/home/regis/projets/overhang/repos/overhang/tutor/tutor/commands/compose.py", line 271, in dc_command
volumes, non_volume_args = bindmounts.parse_volumes(args) File "/home/regis/projets/overhang/repos/overhang/tutor/tutor/bindmounts.py", line 73, in parse_vol
umes
context = custom_docker_compose.make_context("custom", args)
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/click/core.py", line 700, in make_context self.parse_args(ctx, args)
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/click/core.py", line 1045, in parse_args
opts, args, param_order = parser.parse_args(args=args)
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/click/parser.py", line 269, in parse_args
self._process_args_for_options(state)
File "/home/regis/venvs/tutor/lib/python3.8/site-packages/click/parser.py", line 289, in _process_a
rgs_for_options
arg = state.rargs.pop(0)
AttributeError: 'tuple' object has no attribute 'pop'
- [Bugfix] Fix "Read-only file system" errors when running `tutor dev
runserver` commands on Mac OS. (again, see #392)
- [Feature] Automatically bind-mount volumes from the `volumes/` directory with
the `--volume=/...` syntax.
Looks like the `-perm /u=w` argument incorrectly includes files from the
read-only filesystem. This should be fixed by the `-writable` option.
Close #392
This introduces a new dev/local command:
tutor dev bindmount CONTAINER PATH
And a new volume syntax:
tutor dev run --volume=PATH CONTAINER
This syntax automatically bind-mounts folders from the tutorroot/volumes
directory, which is pretty nifty.
- 💥[Improvement] Upgrade Open edX to Koa
- 💥 Setting changes:
- The ``ACTIVATE_HTTPS`` setting was renamed to ``ENABLE_HTTPS``.
- Other ``ACTIVATE_*`` variables were all renamed to ``RUN_*``.
- The ``WEB_PROXY`` setting was removed and ``RUN_CADDY`` was added.
- The ``NGINX_HTTPS_PORT`` setting is deprecated.
- Architectural changes:
- Use Caddy as a web proxy for automated SSL/TLS certificate generation:
- Nginx no longer listens to port 443 for https traffic
- The Caddy configuration file comes with a new ``caddyfile`` patch for much simpler SSL/TLS management.
- Configuration files for web proxies are no longer provided.
- Kubernetes deployment no longer requires setting up a custom Ingress resource or custom manager.
- Gunicorn and Whitenoise are replaced by uwsgi: this increases boostrap performance and makes it no longer necessary to mount media folders in the Nginx container.
- Replace memcached and rabbitmq by redis.
- Additional features:
- Make it possible to disable all plugins at once with ``plugins disable all``.
- Add ``tutor k8s wait`` command to wait for a pod to become ready
- Faster, more reliable static assets with local memory caching
- Deprecation: proxy files for Apache and Nginx are no longer provided out of the box.
- Removed plugin `{{ patch (...) }}` statements:
- "https-create", "k8s-ingress-rules", "k8s-ingress-tls-hosts": these are no longer necessary. Instead, declare your app in the "caddyfile" patch.
- "local-docker-compose-nginx-volumes": this patch was primarily used to serve media assets. The recommended is now to serve assets with uwsgi.
As discussed in
https://discuss.overhang.io/t/lms-worker-timeout-issue/1057/8, the
timeout of the gunicorn worker is a bit low, causing the worker to
crashloop on slower systems. By increasing the timeout to 120 seconds,
the worker should have time to come up before it gets restarted.
- 💥[Improvement] Remove `dev/local pullimages`. Instead, run `dev/local
dc pull`.
- 💥[Improvement] Add `dev dc` and `local dc` commands as thin wrappers
of the `docker-compose` CLI.
- 💥[Improvement] Remove the undocumented `local run_hook` command.
Instead, run `local init --limit=...`.
- 💥[Improvement] Remove `tutor android pullimage` command. Instead, run
`tutor images pull android`.
- [Bugfix] Fix `config save` on Windows.
- [Bugfix] Fix platform checking in user ID function
- [Bugfix] Fix dependency error during `pip install tutor` due to urllib3 incompatibility
- [Bugfix] Fix user ID checking under Windows.
- [Bugfix] Fix template rendering for Windows users.
- [Improvement] Switch to `bcrypt` for htpasswd password generation, for better portability on Windows.
- [Improvement] In the openedx production docker image, add some jitter to the gunicorn worker restart process to prevent all workers from restarting at the same time.
requests depends on urllib3<1.26.0, while urllib3==1.26.0 was just
released. We need to introduce a constraint on urllib3 to avoid the
following error when running `pip install -e`:
pkg_resources.ContextualVersionConflict: (urllib3 1.26.0
(./venv/lib/python3.7/site-packages),
Requirement.parse('urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1'),
{'requests'})
We will be able to remove this constraint once requests becomes
compatible with urllib3==1.26.0. To test this change just run `pip
install -e tutor` and then run a dummy command, such as `tutor local
stop`.
- [Improvement] Mount config and setting files in read-only mode
- 💥[Improvement] Enable unit completion tracking by default.
- [Bugfix] Run `apt update` before `apt install` when installing deps in
the openedx Dockerfile
- [Improvement] Make all commands considerably faster
- [Improvement] Make it easier to override Mongodb connection parameters
- [Bugfix] Add support for .woff and .woff2 font files in themes (thanks @mrtndwrd!)
CORS headers are no longer handled by nginx, but directly by the LMS.
This makes it possible by 3rd-party apps to easily add domain names to
the CORS whitelist, even when they are not subdomains of the LMS.
- [Feature] Add ``images printtag`` command
- [Improvement] Make it possible to override individual contact email
addresses with plugins
- [Bugfix] Replace "no-reply@LMS_BASE" email address by regular contact
email address
- [Bugfix] Disable learner records globally by default
- [Improvement] Upgrade to the latest release of MySQL 5.6
- [Improvement] Non-plugin settings added by "set" directives are now
automatically removed when the plugin is disabled (#241)
- [Improvement] Add CORS basic configuration to LMS for subdomains of
the LMS
- [Feature] Add support for `images build --add-host` option (thanks
@grinderz!)
- [Bugfix] Fix podman compatibility by replacing `docker-compose rm`
command by `docker-compose stop` when stopping containers
- [Improvement] Improve plugin data deletion
- [Improvement] Introduce the `OPENEDX_COMMON_VERSION` setting
- [Bugfix] Make it possible to run init jobs without starting the entire
platform
- [Improvement] Reduce "openedx" Docker image size with static asset
de-duplication
This is done by explicitely listing job dependencies. Unfortunately,
it's not yet possible to move `init` before `start` in `quickstart`,
because some services, such as discovery, depend on the LMS, which takes
a few seconds to boot up. Actually, discovery also depends on ngins, as
it points to "local.overhang.io" when referring to the lms.
- [Bugfix] Upgrade all services to open-release/juniper.3
- [Bugfix] Fix upload of video transcripts to S3
- [Improvement] Memorize whether the user is running a production
platform during interactive configuration
- [Bugfix] Fix incorrect loading of some resources from localhost:18000
in development
- [Bugfix] Fix Samesite=None Secure=False cookie error for users
accessing the LMS with the latest release of Google Chrome
- [Security] Apply javascript security patch ([pull
request](https://github.com/edx/edx-platform/pull/24762))
- [Bugfix] Fix "FileError" on Scorm package upload in Scorm XBlock
- 💥[Improvement] Serve openedx static assets with
[whitenoise](http://whitenoise.evans.io/en/stable/) instead of nginx.
This removes the `k8s-deployments-nginx-init-containers` patch. Plugins
are encouraged to implement static asset serving with Whitenoise as
well.
- [Bugfix] Fix dependency on mysql service when mysql is not activated
- [Improvement] Improve openedx Docker image build time and size with
multi-stage build
- 💥[Feature] Get rid of outdated sysadmin dashboard in LMS at /sysadmin
In the LMS, some resources where loaded from localhost:18000. For
instance: http://localhost:18000/static/images/logo.png
This was due to the fact that the LMS_BASE, LMS_ROOT_URL and thus
SITE_NAME settings are overwritten by the devstack settings, so we need
to define them again in development.
This drastically simplifies volume management, as it is no longer
necessary to manually copy static assets from the docker image to the
bind-mounted volume.
This deprecates the "k8s-deployments-nginx-init-containers" patch, as we
no longer need to init the nginx container. Plugins are encouraged to
start using whitenoise as well for serving static assets.
TODO:
- test media serving: DOES NOT WORK. Whitenoise was designed to serve a
fixed list of static files. Godammit.
- compare performances
This reduces the size of the final image from 3.25Gb to 2.8Gb. Also, it
should be faster to rebuild the image in most cases. For instance, we
will not have to re-install nodejs requirements after part of the
edx-platform repo was modified.
