The manifests we support marshalling/unmarshalling in composer are a
subset of the ones supported by osbuild. osbuild needs to preserve
backwards compatibility, but composer only needs to be concerned with
producing valid manifests at a given time.
A problem with this is that when we marshal our manifests to disk, and
read them out with a future version of composer, we would have to make
sure that the serialization is compatible, or unmarshalling the entire
store will fail. This is what happened when commit 5938276d3b
changed the way we marshal the file source.
Improve on this by only dropping those composes from the store on
unmarshal, whose manifests we are no longer able to unmarshal.
In future commits we will drop that requirement too.
Signed-off-by: Tom Gundersen <teg@jklm.no>
`log` is `nil`, unless we are in verbose mode. Skip logging if no logger
was passed.
Long-term I think we should reconsider how we do our logging, and always
log these type of unexpected errors, and hence always have the logger be
non-nil. For now, fix the immediate problem.
The fact that loading the store from disk fails on upgrade is
unexpected, and should be fixed separately.
Fixes#685.
Signed-off-by: Tom Gundersen <teg@jklm.no>
This is how it is used in the rest of the code, as a name to represent
the repository in the weldr API. Rename to match its use, and avoid
confusion with the ID passed to dnf-json, which is not the same.
Signed-off-by: Tom Gundersen <teg@jklm.no>
The `jobs/:job_id/builds/:build_id/image` route was awkward: the
`:jobid` was actually weldr's compose id and `:build_id` was always `0`.
Change it to `jobs/:job_id/artifacts/:name`, where `:job_id` is now a
job id, and `:name` is the name of the artifact to upload. In the
future, it could support uploading more than one artifact.
This allows removing outputs from `store`, which is now back to being a
pure JSON-store. Take care that `weldr` returns (and deletes) images
from the new (or for backwards compatibility, the old) location.
The `org.osbuild.local` target continues to exist as a marker for the
worker to know whether it should upload artifacts.
GetImageBuildResult() method is not called anywhere, so we can drop it. As
it was the only reader of result.json file, we can also drop all the code
which wrote to that file.
By the same logic as the previous patch.
This exposed an apparent bug, where we are sorting by the version field
of the blueprint in the change objects, but these were never
(un)marshalled, so would always be set to zero in the past.
Remove the no-op sorting for now, as it never had an effect, and
put a note whether we should instantiate the blueprint in the change
object somehow.
Signed-off-by: Tom Gundersen <teg@jklm.no>
Apply the backwards compatibility where the ImageBuild struct is
assembled. This should make the code simpler to reason about, by
reducing the structs we produce and later fix up.
Signed-off-by: Tom Gundersen <teg@jklm.no>
The store only serves the weldr API, and that hard-codes the assumption
of only one image build per compose all over the place. Move this
assumption into the json serialization handler.
Signed-off-by: Tom Gundersen <teg@jklm.no>
Serialize to a plain string, and do the translation for compatibility
from string to string.
This removes a lot of code, and if we change the store format, we should
just skip the conversion altogether.
Signed-off-by: Tom Gundersen <teg@jklm.no>
We may want to move this to the test_distro package in the future, but
calling it something else than its real name is just confusing.
Signed-off-by: Tom Gundersen <teg@jklm.no>
This reduces the amount of resolving and error checking we have to do.
This exposed a bug in weldr's ComposeEntry type, which will be fixed in
a follow-up commit.
Signed-off-by: Tom Gundersen <teg@jklm.no>
The types exposed by the compose package are only used in the store API,
so move them there where they belong.
Making the ownership of the types clear, rather than having them live in
a package for themselves, will make it clearer how the types can be
modified in follow-up commits.
Also remove the JSON annotations, as these types are no longer used for
serialization.
Signed-off-by: Tom Gundersen <teg@jklm.no>
Distinguish between the types used at runtimes and the types used for
(un)marshaling. Also make the types private to the store package.
This should allow us to reduce the interdependencies between the
packages and more easily change things without accidentally breaking
backwards compatibility.
Signed-off-by: Tom Gundersen <teg@jklm.no>
This way we can make more of the store fields and types private in
follow up commits.
This is not a functional change.
Signed-off-by: Tom Gundersen <teg@jklm.no>
After the queue rework, TestCompose() was the only user of
UpdateImageBuildInCompose() and Compose.UpdateState(). Set all required
fields in TestCompose() directly instead of going through those.
The store is responsible for two things: user state and the compose queue. This
is problematic, because the rcm API has slightly different semantics from weldr
and only used the queue part of the store. Also, the store is simply too
complex.
This commit splits the queue part out, using the new jobqueue package in both
the weldr and the rcm package. The queue is saved to a new directory `queue/`.
The weldr package now also has access to a worker server to enqueue and list
jobs. Its store continues to track composes, but the `QueueStatus` for each
compose (and image build) is deprecated. The field in `ImageBuild` is kept for
backwards compatibility for composes which finished before this change, but a
lot of code dealing with it in package compose is dropped.
store.PushCompose() is degraded to storing a new compose. It should probably be
renamed in the future. store.PopJob() is removed.
Job ids are now independent of compose ids. Because of that, the local
target gains ComposeId and ImageBuildId fields, because a worker cannot
infer those from a job anymore. This also necessitates a change in the
worker API: the job routes are changed to expect that instead of a
(compose id, image build id) pair. The route that accepts built images
keeps that pair, because it reports the image back to weldr.
worker.Server() now interacts with a job queue instead of the store. It gains
public functions that allow enqueuing an osbuild job and getting its status,
because only it knows about the specific argument and result types in the job
queue (OSBuildJob and OSBuildJobResult). One oddity remains: it needs to report
an uploaded image to weldr. Do this with a function that's passed in for now,
so that the dependency to the store can be dropped completely.
The rcm API drops its dependencies to package blueprint and store, because it
too interacts only with the worker server now.
Fixes#342
Code that's calling PushCompose() had to depsolve packages and fetch the
right ImageType from a distro, but not create the osbuild manifest. That
was left for PushCompose to do. Move it out of there to the callers, so
that the store is mainly concerned with storing things.
This also simplifies the argument list of PushCompose().
Yes. This goes against my desire not to change code to accommodate
tests. But there is no other good way to test compose results without
long running, and possibly fragile, composes. And this matches lorax's
behavior.
The change adds support for the ?test=1|2 query parameter to the compose
POST, and a new store function - PushTestCompose that handles creating
the fake compose results.
Passing ?test=1 will create a failed compose. Passing ?test=2 will
create a successful compose, but one without any files.
The purpose of these is to be able to test the compose result API
responses like compose/failed, etc.
If you don't set the time it ends up being the default Go time which is
1/1/1 and when you convert that using UnixNano() you get a nice big
negative number (-6795364578871345152) which then eventually shows up in
the queue, finished, and failed output as 'Fri Aug 30 17:47:39 1754'
and since the Time Travel feature is not yet complete this is
impossible.
The fix is to set it when the job is started.
If a user creates a compose with a size of 0, the default image size for
the image type should be used. Also, certain image types have
requirements for the image size. In order to ensure that the proper image
size is stored in the compose object, the compose's ImageBuild object
uses ImageType.Size() to get the correct image size for the image type.
weldr's store is quite complex and handled serialization itself. Move
that part out into a separate package `jsondb`.
This package is more generic than the store needs: it can write an
arbitrary amount of JSON documents while the store only needs one:
state.json. This is in preparation for future work, which introduces a
queue package that builds on top of `jsondb`.
TOML output of the source wasn't using the correct field names or case.
This adds toml tags to the weldr.SourceConfigV0, and store.SourceConfig
structs.
The weldr API already passes custom repositories, so they end up in the
list twice. This is a bit awkward, because weldr passes something to the
store that the store already knows about. It only works this way right
now though, because the rcm API doesn't need these repositories at all.
It expects all repositories to be passed.
If a compose was interrupted by restarting osbuild-composer it should be
set to failed at startup. This was not working because imgBuild is a
temporary variable, the value stored in ImageBuild needs to be modified
directly.
According to the new guidelines in docs/errors.md.
Note that this does not include code that marshals to a writer that
might fail (when a connection drops, for example).
The automatic local target is only needed when accessing the API via
weldr.
In the store, the target was only added when `stateDir` was not `nil`.
This is only used for testing which doesn't exercise the branch in
weldr. Thus, the same check is not needed there.
Rather than having to assume that we only ever produce one
artifact, have each upload target contain the filename it expects
to upload from the osbuild output.
An image file is always explicitly named in the manifest, and we
leave it up to each distro to decide how this is done, but the
convention is to use the same image filename as used when
downloading the image through weldr.
Now make this policy explicit, by quering the distro for the image
name and inserting it into each upload target.
Signed-off-by: Tom Gundersen <teg@jklm.no>
A job's purpose is to build an osbuild manifest and upload the results
somewhere. It should not know about which distro was used to generate
the pipeline.
Workers depended on the distro package in two ways:
1. To set an osbuild `--build-env`. This is not necessary anymore in new
versions of osbuild. More importantly, it was wrong: it passed the
runner from the distro that is being built, instead of one that
matches the host.
This patch simply removes that logic.
2. To fetch the output filename with `Distro.FilenameFromType()`. While
that is useful, I don't think it warrants the dependency.
This patch uses the fact that all current pipelines output exactly
one file and uploads that. This should probably be extended in the
future to upload all output files, or to name them explicitly in the
upload target.
The worker should now compile to a smaller binary and do less
unnecessary work on startup (like reading repository files).
Now that `Store.PushCompose()` takes a `Distro` as argument, the rcm API
can use that function as well. This moves them both through the same
code path, reducing duplication.
Remove `PushComposeRequest()` and the corresponding struct. It was
supposed to allow composes with multiple output types and architectures,
but that was not yet implemented. Merging the two now simplifies moving
the compose queue out of the store in a future commit, which will then
tackle multi-image-type composes as well.
Only the weldr API has the concept of a default distro. Pass that distro
explicitly to `PushCompose()` and fetch the distro from the compose in
all other functions that accessed Store.Distro.
In the post-dnf-stage world, `Distro.Manifest` expects the full list of
depsolved packages. This is similar to what weldr does, but much
simpler, because the rcm API only cares about base packages.
