By managing volumes in a better fashion and using code that is closer
to being idempotent, while being declarative, we achieve an image that
is closer to the original one, but gets the plugins that we want and
the configuration that we want for integration with DD.
Closes#9. This image now allows for BBB_HOST and BBB_API_SECRET as
variables in dd.conf, which also configure the corresponding plugin on
Nextcloud.
By executing tests in this normalised fashion it is easier to compare
results between different instances or different patch levels.
Roughly speaking there are two kinds of tests:
- VM tests: which summarise general performance of the instance,
without taking DD into account
- DD tests: which simulates many logins and interactions with DD,
while recording the session as a user would pereceive it from a
browser
By using these we should be able to consistently compare and improve
performance.
The original dd-stress-test.tpl.jmx tests file was prepared by
Teradisk with hardcoded instance, threadcount and duration values.
Testing should now be performed with `vm-test.sh` and `dd-test.sh`
respectively, and the template file should stay generic.
Flask-SocketIO depends on dnspython but dnspython 2.3 removes
dns.rdtypes.ANY, which is needed by Flask-SocketIO so we keep it below
version 2.3
We had missed Flask-SocketIO being a dependency on API when fixing the
issue for admin.
By having the environment explicit on each service, we both document
the settings and have more control over what each service is allowed
to see.
This avoids weird things like nginx having access to postgresql's
credentials on its environment.
As a bonus: we are able to use one single environment file, which is
basically dd.conf with some values that are dynamically-calculated and
added from dd-ctl.
That issue is fixed in NC 25, but it will likely not be backported to
NC 24.
It produces issues when modifying users and not modifying their
display name.
See also: https://github.com/nextcloud/server/issues/33751
This is used to tag Docker images in the registry / pull images form
the registry, it has been observed in the wiled that:
git rev-parse --short
can have different default values for its length depending on the
system.
We currently specify the length to be 8 as specified here:
https://git-scm.com/docs/git-rev-parse#Documentation/git-rev-parse.txt---shortlength
The first example of these changes relates to the latest Nextcloud
upgrades to major version 24.
Operators are now expected to read the latest version this file before
updating their instances.
This may be causing issues when the plugin requires occ upgrade
Alternatively, installing plugins may require it.
While there, also remove one of the nextcloud_scan calls and delay it
along with logo customisation until after a potential ./occ upgrade
has taken place.
NotaBLE és la col·laboració entre Gwido i el Workspace educatiu DD.
És un projecte de Xnet, IsardVDI, Gwido i Taller de Músics, guanyador
de la Ciutat Proactiva 2021, suport a la innovació urbana de la
Fundació BitHabitat.
The API spec file can be generated with:
python -m admin.views.test.test_ApiViews --generate-spec
From the admin development environment.
A simple testing ground that serves the Swagger UI can also be started with:
python -m admin.views.test.test_ApiViews
We do this more reliably on HAProxy, as doing it from WP requires
specialised plugins and in DD we are sure that traffic goes through
the corresponding HAProxy backend.
This fixes several issues where services would see the internal IP of
the proxy and not that of the client.
It works by first unsetting any proxy-related headers that arrive from
the internet, then setting those as seen by HAProxy's entrypoint
frontend.
And finally making sure that neither WAF when enabled nor other
HAProxy backends touch these headers, while they are actually used by
the final services.
Services affected: Netcloud, Keycloak, Moodle