Workers already update their state when it changes, and this state
propagates through OBS proxies, so we can safely return it immediately
when requested by the warden, and schedule an update in the background.

This should mitigate timeouts when the warden prods OBS proxy server
prods OBS proxy client prods worker, and while the reply climbs back
through the interpipes, warden times out and kills the job.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

When OBS is told a worker is 'building' it thinks it’s "building for
someone else" and marks in as "away". See bs_repserver for more details.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

This was causing duplicate log lines.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

With recent versions of obs-proxy the client.proxy setting is no longer
recognized and there are sane defaults for client.{host,port} so drop
those all out of the configmap template

Signed-off-by: Sjoerd Simons <sjoerd@collabora.com>

To make it easier to reproduce a setup or roll back to a previous
state, tag each pipeline run with a unique tag of its own in the
addition to a per-branch tag.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Without this processors, HTTP access logs look like this:

    [info] 127.0.0.1:44936 PUT /build 1.1 404 9 3170 [http.server]

With the change applied, they become more informative:

    [info] 127.0.0.1:44936 PUT /build 1.1 404 9 3170 [http.server] http_port=6040 known_worker=False worker_id=None

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

On the contrary, we don’t care about logging ports for WebSockets.

It would be a bit more efficient to store the detected port into
g.port here and reuse it later in require_auth(), but OTOH that
would make the tests for require_auth a bit more complicated, since
before_request decorators are not automatically imported.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Internally-accessible resources (emulating the interface of workers) are
not affected at all by the fact the external interface is prefixed,
so the app overall cannot have root_path set, otherwise communications
between repserver and emulated workers break.

Instead, split out externally-accessible resources into a blueprint,
and mount it into the correct place before starting the webserver.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Python 3.11 is now in Bookworm, so make failures on 3.11 hard failures.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

request.path includes the root prefix, so when it’s not empty, it needs
to be stripped from request.path before using it.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

The fuse-overlayfs configuration is no longer necessary.
However, binfmt-support package was missing from the dependencies.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Drop unused imports, drop extra empty lines, replace a wildcard
import with a series of specific imports.

Fixes: abd77e4d ("Add configuration de/serialisation code")

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Pipe all logs through structlog, including stdlib logs.
When run on a tty, use colours, otherwise logfmt.
Convert some log prints to structured logs.
Bump Quart dependency to 0.18.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

This is necessary to enable the client to be able to only specify
minimal configuration, and fetch the rest from the server.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Andrej Shadura authored 2 years ago and Sjoerd Simons committed 1 year ago

This is an internal endpoint mostly useful for debugging, it should
not be open to the world.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

This enables other tests to override parts of the config without
having to do the manual work themselves.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

When deploying in a proxy-per-worker configuration, it’s useful to
consolidate all proxy instances on a single hostname, since DNS names
can be scarce.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

On lightweight runners, tests would sometimes run out of disk space
or memory or both.

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>