by adding the context from main and make auth.GetAdminContext accepting a
context we are now able to stop the metrics collection loop once the
context is canceled
Signed-off-by: Mario Constanti <mario.constanti@mercedes-benz.com>
refactoring is needed to make the metrics package usable from within the
runner package for further metrics.
This change also makes the metric-collector independent from requests to
the /metrics endpoint
Signed-off-by: Mario Constanti <mario.constanti@mercedes-benz.com>
Without '-linkmode external' some of the produced binaries were not actually
static. /bin/garm was fine but providers would fail with "exit code 127" when
executed. Unpacking the container image and running the binaries in strace
showed that they were accessing the musl dynamic linker.
Signed-off-by: Jeremi Piotrowski <jpiotrowski@microsoft.com>
This change invalidates tokens based on more parameters. Tokens that were
generated for previous attempts of spinning up an instance will be invalidates.
Also, only instances that are in Running or Creating will be able to authenticate.
Signed-off-by: Gabriel Adrian Samfira <gsamfira@cloudbasesolutions.com>
This change increases the tools refresh interval to 5 minutes, cleans
up the websocket code a bit, augments the error message that may be returned
when trying to delete a runner in an invalid state and removes a log message
that does not add much value.
Signed-off-by: Gabriel Adrian Samfira <gsamfira@cloudbasesolutions.com>
I accidentally disabled the log streamer when I moved the config options
to their own section. This change fixes that.
This change also adds some safety checks and locking when cleaning up stale
clients. The websocket hub Write() function now copies the message before
sending it on the channel to the clients.
Signed-off-by: Gabriel Adrian Samfira <gsamfira@cloudbasesolutions.com>
Pool managers will have 2 fields identifying which manager generated
the log line.
In the future, we will add tracking ids in various cases, allowing
us to track down issues faster.
Signed-off-by: Gabriel Adrian Samfira <gsamfira@cloudbasesolutions.com>
This change switches GARM to the new structured logging standard
library. This will allow us to set log levels and reduce some of
the log spam.
Given that we introduced new knobs to tweak logging, the number of
config options for logging now warrants it's own section.
Signed-off-by: Gabriel Adrian Samfira <gsamfira@cloudbasesolutions.com>
Allow runners to update their own system information. Runners can now send
back os_name, os_version and agent_id back as part of a POST to
CALLBACK_URL/system-info/.
The goal is to get better info in regard to the actual OS that's running
and to move the agent_id from the status updates to the system-info callback.
The status updates should be used only to send back info about the status of
the installation process.
Signed-off-by: Gabriel Adrian Samfira <gsamfira@cloudbasesolutions.com>
Canonical have relicensed the LXD project to AGPLv3. This means that we can
no longer update the go LXD client without re-licensing GARM as AGPLv3. This
is not desirable or possible.
The existing code seems to be Apache 2.0 and all code that has already been
contributed seems to stay as Apache 2.0, but new contributions from Canonical
employees will be AGPLv3.
We cannot risc including AGPLv3 code now or in the future, so we will separate
the LXD provider into its own project which can be AGPLv3. GARM will simply
execute the external provider.
If the client code of LXD will ever be split from the main project and re-licensed
as Apache 2.0 or a compatible license, we will reconsider adding it back as a
native provider. Although in the long run, I believe external providers will
be the only option as they are easier to write, easier to maintain and safer to
ship (a bug in the provider does not crash GARM itself).
Signed-off-by: Gabriel Adrian Samfira <gsamfira@cloudbasesolutions.com>
