Introduce a new `ComputeImageInsert()` method for importing images into
GCP. It uses the `compute.Images.Insert()` API [1], which has many
advantages over the currently used way of importing images using the
CloudBuild API. The advantages are mainly that the image is imported as
is and no additional cache files or VMs are created as part of the
import process. Therefore there is no need to do additional cleanup of
cache files after importing the image.
In addition, the import itself is approximately 30% faster for RHEL
images when using the `Insert()` call.
Nevertheless the `Insert()` call accepts only gzip-ed tarball with a RAW
disk, unlike the `Import()` call, which accepts basically any virtual
disk format.
[1] https://cloud.google.com/compute/docs/reference/rest/v1/images/insert
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Add the `gce-rhui` image type intended for Google Compute Engine. The image
uses Google's RHUI infrastructure to access Red Hat content.
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Add the `gce` image type intended for Google Compute Engine. The image
is BYOS - bring your own subscription and requires registering in order
to access Red Hat content.
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Extend RHEL-84 `imageTypeS2` structure to contain pipelines generator
function. Previously, the `imageTypeS2` implementation defaulted to only
a single pipelines generator method for EDGE image types. The ability to
pass a different generator function implementation is important to
enable addition of new image types relying on osbuild Manifest v2.
Rename the original pipeline generator method to `edgePipelines`.
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Add the `gce-rhui` image type intended for Google Compute Engine. The image
uses Google's RHUI infrastructure to access Red Hat content.
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Add the `gce` image type intended for Google Compute Engine. The image
is BYOS - bring your own subscription and requires registering in order
to access Red Hat content.
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Add the `gce-rhui` image type intended for Google Compute Engine. The image
uses Google's RHUI infrastructure to access Red Hat content.
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Add the `gce` image type intended for Google Compute Engine. The image
is BYOS - bring your own subscription and requires registering in order
to access Red Hat content.
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Add the `gce` image type intended for Google Compute Engine. The image
is BYOS - bring your own subscription and requires registering in order
to access Red Hat content.
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Introduce `tarArchivePipeline()` function returning a pipeline,
which creates a Tar archive from another pipeline tree referenced by the
pipeline name.
Replace `tarStage()` with `osbuild.NewTarStage()`
Use the `tarArchivePipeline()` function in respective image type
pipelines.
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Introduce `tarArchivePipeline()` function returning a pipeline, which
creates a Tar archive from another pipeline tree referenced by the
pipeline name.
Replace `tarStage()` with `osbuild.NewTarStage()`
Use the `tarArchivePipeline()` function in respective image type
pipelines.
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Introduce `tarArchivePipeline()` function returning a pipeline, which
creates a Tar archive from another pipeline tree referenced by the
pipeline name.
Replace `tarStage()` with `osbuild.NewTarStage()`
Use the `tarArchivePipeline()` function in respective image type
pipelines.
Signed-off-by: Tomas Hozza <thozza@redhat.com>
The `tar` stage options contain three boolean values. All of them
default to `true` in the osbuild stage implementation [1]. However
if these values were explicitly set to `false`, they would be omitted
from the resulting JSON structure. As a result, it was impossible to use
any non-default values.
Use `*bool` instead of `bool`, to ensure that explicitly set `false`
values will end up in the JSON structure passed to osbuild.
[1] 8102f20d23/stages/org.osbuild.tar (L39-L53)
Signed-off-by: Tomas Hozza <thozza@redhat.com>
Add depsolve job error dependency test cases for
regular composes and koji composes. The error furthest
up the chain should be returned in the details field
of the job error.
If an osbuild or koji-osbuild job has failed, add
a check to see if it is a result of the build jobs
dependencies and return the dependency failure job
error furthest up the chain of errors & add this
error to the details filed of the build job error.
Add a helper function to query dependency
failures of osbuild & koji-osbuild jobs.
If a build job has a dependency error the
function will check for the job error of the
manifest job. If that also has a dependency
error the function will query the depsolve
job too for a job error.
Add a helper function to check for dependency
errors for job errors. This simply returns true
if a job error has a dependency error code and
false otherwise.
Specify a size for the root filesystem in the partition table,
which basically equates to a minimum size. In reality all image
types specify a larger image size and thus we enlarge the root
file system to more than the specified size for plain layouts.
Does not change any existing manifests.
This also prepares the enablement of auto-LVM conversion, since
in that case we need to have a size for the root file system
specified.
jobimpl-osbuild
---------------
Add GenericS3Creds to struct
Add method to create AWS with Endpoint for Generic S3 (with its own credentials file)
Move uploading to S3 and result handling to a separate method (along with the special VMDK handling)
adjust the AWS S3 case to the new method
Implement a new case for uploading to a generic S3 service
awscloud
--------
Add wrapper methods for endpoint support
Set the endpoint to the AWS session
Set s3ForcePathStyle to true if endpoint was set
Target
------
Define a new target type for the GenericS3Target and Options
Handle unmarshaling of the target options and result for the Generic S3
Weldr
-----
Add support for only uploading to AWS S3
Define new structures for AWS S3 and Generic S3 (based on AWS S3)
Handle unmarshaling of the providers settings' upload settings
main
----
Add a section in the main config for the Generic S3 service for credentials
If provided pass the credentials file name to the osbuild job implementation
Upload Utility
--------------
Add upload-generic-s3 utility
Makefile
------
Do not fail if the bin directory already exists
Tests
-----
Add test cases for both AWS and a generic S3 server
Add a generic s3_test.sh file for both test cases and add it to the tests RPM spec
Adjust the libvirt test case script to support already created images
GitLabCI - Extend the libvirt test case to include the two new tests
De-duplicate the `qemuStageOptions()` implementation from many distro
implementations and move it to the `osbuild2` package as
`NewQEMUStageOptions()`.
Adjust all affected pipeline implementations.
The `NewQEMUStageOptions()` function takes an optional format-specific
options argument, instead of a dedicated format-specific options (the
previously used qcow2 compat version). This is a preparation to produce
VMDK images with a different than default subformat.
Add unit test for `NewQEMUStageOptions()`.
Support setting a specific VMDK subformat in the type format options.
The required osbuild version in the SPEC file is not bumped, since the
new functionality is currently not used by any image type.
Related to https://github.com/osbuild/osbuild/pull/999
Define a new aliased type `QEMUFormat` for the format type used by the
osbuild2 QEMU stage and define constants for all allowed values.
Use QEMU format type constants in all relevant places, instead of string
literals. Not using string literals minimizes the room for making a
typo.
We want to ensure that cloud images connect to Red Hat[1] independently
of how the content was acquired (PAYG, BYOS, or marketplace).
This auto-registration feature is already enabled for AWS and this
patch enables it for Azure with the same recommended settings:
Services:
rhsmcertd: Enabled (already done, so not changed in the patch)
/etc/rhsm/rhsm.conf:
auto_registration: enabled
auto_registration_interval: 60 (the default, so not explicitly set)
manage_repos: false
The latter value `manage_repos` is left enabled (the default) in case
the user explicitly requested to have the system subscribed, i.e. the
`RHSMConfigWithSubscription` code path.
Regenerate the relevant test manifests and image information.
[1] https://cloud.redhat.com
[2] https://docs.google.com/document/d/1VeZFJxNUlyZMQJh6s3NA3RLvadqATsGxVet6uuP87_4
The `ManifestJobStatus` and `ManifestByIdJobStatus` both
had identical functionality. The `ManifestByIdJobStatus`
is not being referenced anywhere in the codebase and so
this function has been removed.
The image type is only ever known (externally) as image-installer.
Renaming the internal variables and functions to reflect the name makes
the code easier to navigate.
The users anaconda module enables users to create user accounts at
install time if one is not already created in the payload. This is
required for the cloud API (Image Builder service) for the image
installer where user customizations are not supported. Without it, user
creation isn't possible on the installed system.
The module also enables user creation at install time through the
kickstart file for both the image-installer and the edge-installer
(Anaconda only).
Therefore, for the image-installer, the users and groups are no longer
created as part of the payload.
This commit adapts the changes from the following commits (originally
made in the RHEL 8.6 and RHEL 9.0 distros) to the rest of the RHEL
distro definitions:
ebc3330cbd5825294dad
Use single NewAnacondaStageOptions() from osbuild2 instead of
implementing in each distro.
Followup from 974b0a434b, copied to the
rest of the RHEL distro definitions.
Use single NewKickstartStageOptions() and replace image-type-specific
implementation from each distro.
- Followup from cb186df208, copied to the
rest of the RHEL distro definitions.
NB: The change was not made in the Fedora distro definitions as they are
currently being rewritten.
Use single NewGroupsStageOptions() from osbuild1 and osbuild2 instead of
implementing in each distro.
- Followup from 2eef6e6e2d, copied to the
rest of the RHEL distro definitions.
- Added NewGroupsStageOptions() to osbuild1 for rhel8 and rhel84.
NB: The change was not made in the Fedora distro definitions as they are
currently being rewritten.
Use single NewUsersStageOptions() from osbuild1 and osbuild2 instead of
implementing in each distro.
- Followup from ca8b371142, copied to the
rest of the RHEL distro definitions.
- Added NewUsersStageOptions() to osbuild1 for rhel8 and rhel84.
NB: The change was not made in the Fedora distro definitions as they are
currently being rewritten.
Followup from, f34380d5b5 and
3a1765a5a8, copied to the rest of the RHEL
distro definitions.
For now, these customizations have no effect on the manifest.
The new `with-users` variants of the edge-installer test cases include
the user customizations in the blueprint, but the manifests are
(currently) the same as the corresponding base cases.
All dependency errors, whether they are 4xx or 5xx,
are currently being considered as a 5xx error in parent
jobs. This is causing some of the build alerts to fire
off when a depsolve job has failed, for example, when
in reality, this is an expected result. This commit
ensures that dependency errors are being reported as
4xx status in monitoring.
