Before edx-platform was upgraded to Celery 5, lms-worker and
cms-worker could be invoked using this syntax:
celery worker --app=APP <args> --maxtasksperchild=N <args>
Since the recent Celery 5 upgrade (edx-platform commit 0588c92),
though, this fails with the messages:
You are using `--app` as an option of the worker sub-command:
celery worker --app celeryapp <...>
The support for this usage was removed in Celery 5.0.
Instead you should use `--app` as a global option:
celery --app celeryapp worker <...>
and:
Error: No such option: --maxtasksperchild
(Possible options: --max-memory-per-child, --max-tasks-per-child)
So, this commit changes the lms-worker and cms-worker invocations to:
celery --app=APP <args> --max-tasks-per-child=N <args>
In the LMS/CMS Dockerfile, the env var STUDIO_CFG is set
in order to point CMS at its configuration json/yaml file.
Since https://github.com/edx/edx-platform/pull/29534
(which introduced 0013-cms-vs-studio.rst), the STUDIO_CFG
variable has been deprecated in favor of CMS_CFG.
This change updates the Dockerfile to reflect the new
preferred environment variable.
The only noticeable impact of this change is that it
will remove a depreation warning from Django startup
for tutor uses running off of Open edX master.
This introduces quite a few changes to make it easier to run Caddy as a load
balancer in Kubernetes:
- Make it possible to start/stop a selection of resources with ``tutor k8s
start/stop [names...]``.
- Make it easy to deploy an independent LoadBalancer by converting the caddy
service to a NodePort when ``ENABLE_WEB_PROXY=false``.
- Add a ``app.kubernetes.io/component: loadbalancer`` label to the LoadBalancer
service.
- Add ``app.kubernetes.io/name`` labels to all services.
- Preserve the LoadBalancer service in ``tutor k8s stop`` commands.
- Wait for the caddy deployment to be ready before running initialisation jobs.
Close #532.
On some providers (notably: DigitalOcean) NodePort services are not exposed to
the outside world. But this is not what the Kubernetes spec describes:
https://kubernetes.io/docs/concepts/services-networking/service/#publishing-services-service-types
Thus, there is a risk that NodePort services are exposed to the outside world
in some context. To avoid this, we convert all NodePort to ClusterIP resources.
Python 3.5 has reached end of life in September 3.5. Anyway, Tutor was not
compatible because some dev dependencies, such as astroid 2.8.3, are no longer
available in 3.5.
This means that we can now start using many python 3.6 niceties, such as
f-strings \o/
Through the commonLabels directive in kustomization.yml, all resources
get a label named "app.kubernetes.io/version", which is being set to
the Tutor version at the time of initial deployment.
When the user then subsequently progresses to a new Tutor version,
Kubernetes attempts to update this label — but for Deployment,
ReplicaSet, and DaemonSet resources, this is no longer allowed as of
https://github.com/kubernetes/kubernetes/issues/50808. This causes
"tutor k8s start" (at the "kubectl apply --kustomize" step) to break
with errors such as:
Deployment.apps "redis" is invalid: spec.selector: Invalid value: v1.LabelSelector{MatchLabels:map[string]string{"app.kubernetes.io/instance":"openedx-JIONBLbtByCGUYgHgr4tDWu1", "app.kubernetes.io/managed-by":"tutor", "app.kubernetes.io/name":"redis", "app.kubernetes.io/part-of":"openedx", "app.kubernetes.io/version":"12.1.7"}, MatchExpressions:[]v1.LabelSelectorRequirement(nil)}: field is immutable
Simply removing the app.kubernetes.io/version label from
kustomization.yml will permanently fix this issue for newly created
Kubernetes deployments, which will "survive" any future Tutor version
changes thereafter.
However, *existing* production Open edX deployments will need to throw
the affected Deployments away, and re-create them.
Also, add the Tutor version as a resource annotation instead, using
the commonAnnotations directive.
See also:
https://github.com/kubernetes/client-go/issues/508https://kubectl.docs.kubernetes.io/references/kustomize/kustomization/commonlabels/https://kubectl.docs.kubernetes.io/references/kustomize/kustomization/commonannotations/
Fixes #531.
When nginx was removed in favour of caddy, we decided that plugin
implementations of the "caddyfile" patch should make use of the "port" local
variable. However, local variables are not available from inside plugin
patches, which are rendered outside of the context of the parent templates.
For a more extensive description of the problem, see:
https://github.com/overhangio/tutor-mfe/pull/23#issuecomment-964016190
We still want to make it easy for developers to decide what should the port be
for caddy hosts. To do so, we make use of environment variables that are passed
at runtime to the caddy container.
Thus, a regular plugin patch should look like this:
{{ PLUGIN_HOST }}{$default_site_port} {
import proxy "myplugin:8000"
}
Forum is an optional feature, and as such it deserves its own plugin. Starting
from Maple, users will be able to install the forum from
https://github.com/overhangio/tutor-forum/
Close #450.
In the past, tutor was installed with "pip install tutor-openedx". For
some time (since v12.0.2), "tutor" was installed as a dependency of
"tutor-openedx". Now is the time to get rid of that old package.
The standard way of installing tutor is now with "pip install tutor".
With this change, containers are no longer run as "root" but as unprivileged
users. This is necessary in some environments, notably some Kubernetes
clusters.
To make this possible, we need to manually fix bind-mounted volumes in
docker-compose. This is pretty much equivalent to the behaviour in Kubernetes,
where permissions are fixed at runtime if the volume owner is incorrect. Thus,
we have a consistent behaviour between docker-compose and Kubernetes.
We achieve this by bind-mounting some repos inside "*-permissions" services.
These services run as root user on docker-compose and will fix the required
permissions, as per build/permissions/setowner.sh These services simply do not
run on Kubernetes, where we don't rely on bind-mounted volumes. There, we make
use of Kubernete's built-in volume ownership feature.
With this change, we get rid of the "openedx-dev" Docker image, in the sense
that it no longer has its own Dockerfile. Instead, the dev image is now simply
a different target in the multi-layer openedx Docker image. This makes it much
faster to build the openedx-dev image.
Because we declare the APP_USER_ID in the dev/docker-compose.yml file, we need
to pass the user ID from the host there. The only way to achieve that is with a
tutor config variable. The downside of this approach is that the
dev/docker-compose.yml file is no longer portable from one machine to the next.
We consider that this is not such a big issue, as it affects the development
environment only.
We take this opportunity to replace the base image of the "forum" image. There
is now no need to re-install ruby inside the image. The total image size is
only decreased by 10%, but re-building the image is faster.
In order to run the smtp service as non-root, we switch from namshi/smtp to
devture/exim-relay. This change should be backward-compatible.
Note that the nginx container remains privileged. We could switch to
nginxinc/nginx-unprivileged, but it's probably not worth the effort, as we are
considering to get rid of the nginx container altogether.
Close #323.
