* fix: make config read/write thread safe and fix some other similar issues
1. The config config has a lock, and safe methods to update and read the attributes
2. The config has methods to retrieve copies of specific attributes, such as the extyensions config, the auth config, and the authz config.
These are needed, as the config object may mutate in the middle of an auth/authz requests, and we avoid partial configuration being applied for that request.
3. Fix an issue with the monitoring server not stopping when the controller is shut down.
4. Fix an issue with the HTPasswdWatcher not stopping when the background tasks are supposed to finish.
5. Fix some tests using hardcoded ports.
Moved some of the methods which were on the main config to the auth, access control and extension configs
Signed-off-by: Andrei Aaron <andreifdaaron@gmail.com>
* fix: migrate to Go module v2 for proper semantic versioning
This change updates the module path from 'zotregistry.dev/zot' to
'zotregistry.dev/zot/v2' to comply with Go's semantic versioning rules.
According to Go's module versioning requirements, major version v2+
must include the major version in the module path. The current
module path 'zotregistry.dev/zot' only supports v0.x.x and v1.x.x
versions, making existing v2.x.x tags (like v2.1.8) unusable.
Changes:
- Updated go.mod module path to zotregistry.dev/zot/v2
- Updated all internal import paths across 280+ Go source files
- Updated configuration files (golangcilint.yaml, gqlgen.yml)
- Updated README.md Go reference badge
This fix enables proper use of existing v2.x.x Git tags and allows
external packages to import zot v2+ versions without compatibility
errors.
Resolves: Go module import compatibility for v2+ versions
Fixes: #3071
Signed-off-by: Luca Muscariello <muscariello@ieee.org>
* fix: regenerate GraphQL files with updated v2 import paths
The gqlgen tool needs to regenerate the GraphQL schema files after
the module path change to use the new v2 imports.
Signed-off-by: Luca Muscariello <muscariello@ieee.org>
---------
Signed-off-by: Luca Muscariello <muscariello@ieee.org>
Most users don't make the difference between retention deleting untagged manifests vs GC deleting other blobs.
This causes confusion since the GC delay and the retention delay (used for untagged manifests and orphan referrers) have different defaults, and are set separately in the zot configuration.
Most users don't configrue retention policies, and they still expect untagged manifests to be deleted at GC time.
With this change, if retention delay is not specified in the config file, the value used is the GC delay.
If GC delay is also unspecified in the config file, the default GC delay is used for both.
Signed-off-by: Andrei Aaron <andreifdaaron@gmail.com>
* fix: migrate from github.com/rs/zerolog to golang-native log/slog
We have been using zerolog for a really long time.
golang now has structured logging using slog.
Best to move to this in interests of long-term support.
This is a tech debt item.
Signed-off-by: Ramkumar Chinchani <rchincha.dev@gmail.com>
* fix: a few changes on top
Signed-off-by: Ramkumar Chinchani <rchincha.dev@gmail.com>
* fix: address comments
Signed-off-by: Ramkumar Chinchani <rchincha.dev@gmail.com>
---------
Signed-off-by: Ramkumar Chinchani <rchincha.dev@gmail.com>
* chore: increase/stabilize coverage for the local storage driver
Signed-off-by: Andrei Aaron <andreifdaaron@gmail.com>
* chore: add/stabilize coverage for soring ImageSummary objects
Signed-off-by: Andrei Aaron <andreifdaaron@gmail.com>
* chore: stabilize coverage in sync tests
Signed-off-by: Andrei Aaron <andreifdaaron@gmail.com>
---------
Signed-off-by: Andrei Aaron <andreifdaaron@gmail.com>
- fixes#3347: removeUntaggedManifests() did not consider compatible manifest types
- add AsDockerImage() to Image and MultiarchImage for testing
- extend TestGarbageCollectAndRetentionMetaDB to test docker image and multiarch image
Signed-off-by: Stephan Merker <stephan.merker@sap.com>
Using just the last repository is not enough as in the case when it is deleted
(either by GC or some other way), GetNextRepository returns empty string
causing the generator to be marked completed without any errors.
An alternative would have been to start over from the first repository,
but this can take hours if multiple repositories need to be deleted,
not to mention the processing power and I/O and S3 load this could take.
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* chore: bump github.com/olekukonko/tablewriter from 0.0.5 to 1.0.7
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* fix: zli failed to connect to https server using test certificates
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
---------
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
It is to fix#3185.
This fixes the case where MetaDB is not instantiated (none of the conditions match),
and we want to retain tags only by pattern (which should not need to use MetaBD).
Without this fix you could only use retention to delete untagged manifests.
If you specified only the key "patterns" under "keepTags", zot would crash.
It was possible to not specify "keepTags" all, which would retain all tags,
but it was not possible to retains specific tags.
Basically the case quoted below, from the documentation, was broken::
https://zotregistry.dev/v2.1.4/articles/retention/#configuration-example
```
When you specify a regex pattern with no rules other than the default, all tags matching the pattern are retained.
```
This would only work if MetaDb was instantiated by an unrelated configured feature.
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
Fixes#3005
Previously, changing a image's media-type was disallowed.
However, "docker buildx" appears to first push an image manifest and
then an image index for the same image tag. So, allow this.
Signed-off-by: Ramkumar Chinchani <rchincha.dev@gmail.com>
* feat: add redis cache support
https://github.com/project-zot/zot/pull/2005
Fixes https://github.com/project-zot/zot/issues/2004
* feat: add redis cache support
Currently, we have dynamoDB as the remote shared cache but ideal only
for the cloud use case.
For on-prem use case, add support for redis.
Signed-off-by: Ramkumar Chinchani <rchincha@cisco.com>
* feat(redis): added blackbox tests for redis
Signed-off-by: Petu Eusebiu <peusebiu@cisco.com>
* feat(redis): dummy implementation of MetaDB interface for redis cache
Signed-off-by: Alexei Dodon <adodon@cisco.com>
* feat: check validity of driver configuration on metadb instantiation
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat: multiple fixes for redis cache driver implementation
- add missing method GetAllBlobs
- add redis cache tests, with and without mocking
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat(redis): redis implementation for MetaDB
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat(redis): use redsync to block concurrent write access to the redis DB
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat(redis): update .github/workflows/cluster.yaml to also test redis
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat(metadb): add keyPrefix parameter for redis and remove unneeded method meta.Crate()
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat(redis): support RedisCluster configuration and add unit tests
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat(redis): more tests for redis metadb implementation
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat(redis): add more examples and update examples/README.md
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat(redis): move option parsing and redis client initialization under pkg/api/config/redis
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* chore(cachedb): move Cache interface to pkg/storage/types
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat(redis): reorganize code in pkg/storage/cache.go
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat(redis): call redis.SetLogger() with the zot logger as parameter
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
* feat(redis): rename pkg/meta/redisdb to pkg/meta/redis
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
---------
Signed-off-by: Ramkumar Chinchani <rchincha@cisco.com>
Signed-off-by: Petu Eusebiu <peusebiu@cisco.com>
Signed-off-by: Alexei Dodon <adodon@cisco.com>
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
Co-authored-by: a <a@tuxpa.in>
Co-authored-by: Ramkumar Chinchani <rchincha@cisco.com>
Co-authored-by: Petu Eusebiu <peusebiu@cisco.com>
Co-authored-by: Alexei Dodon <adodon@cisco.com>
* feat: add support for docker images
Issue #724
A new config section under "HTTP" called "Compat" is added which
currently takes a list of possible compatible legacy media-types.
https://github.com/opencontainers/image-spec/blob/main/media-types.md#compatibility-matrix
Only "docker2s2" (Docker Manifest V2 Schema V2) is currently supported.
Garbage collection also needs to be made aware of non-OCI compatible
layer types.
feat: add cve support for non-OCI compatible layer types
Signed-off-by: Ramkumar Chinchani <rchincha@cisco.com>
*
Signed-off-by: Ramkumar Chinchani <rchincha@cisco.com>
* test: add more docker compat tests
Signed-off-by: Ramkumar Chinchani <rchincha@cisco.com>
* feat: add additional validation checks for non-OCI images
Signed-off-by: Ramkumar Chinchani <rchincha@cisco.com>
* ci: make "full" images docker-compatible
Signed-off-by: Ramkumar Chinchani <rchincha@cisco.com>
---------
Signed-off-by: Ramkumar Chinchani <rchincha@cisco.com>
Exit early in case of all folders and known non-blob file names.
This avoids the logic for validating digests, and log message generation.
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
Looks like we didn't have many GC tests for retaining multiarch images.
I added more data to the existing image retention tests, besides the new GC tests.
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
cache.HasBlob() looks in both buckets: duplicates and original blobs
Because we want to check if the blob is in original bucket let's use
cache.GetBlob() because it's looking only in original bucket.
Signed-off-by: Eusebiu Petu <petu.eusebiu@gmail.com>
In case of delete by tag only the tag is removed, the manifest itself would continue to be accessible by digest.
In case of delete by digest the manifest would be completely removed (provided it is not used by an index or another reference).
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
when a client pushes an image zot's inline dedupe
will try to find the blob path corresponding with the blob digest
that it's currently pushed and if it's found in the cache
then zot will make a symbolic link to that cache entry and report
to the client that the blob already exists on the location.
Before this patch authorization was not applied on this process meaning
that a user could copy blobs without having permissions on the source repo.
Added a rule which says that the client should have read permissions on the source repo
before deduping, otherwise just Stat() the blob and return the corresponding status code.
Signed-off-by: Petu Eusebiu <peusebiu@cisco.com>
Co-authored-by: Petu Eusebiu <peusebiu@cisco.com>
* fix(oras)!: remove ORAS artifact references support
ORAS artifacts/references predated OCI dist-spec 1.1.0 which now has the
same functionality and likely to see wider adoption.
Signed-off-by: Ramkumar Chinchani <rchincha@cisco.com>
* test: update to released official images
So that they are unlikely to be deleted.
*-rc images may be cleaned up over time.
Signed-off-by: Ramkumar Chinchani <rchincha@cisco.com>
---------
Signed-off-by: Ramkumar Chinchani <rchincha@cisco.com>
This causes the "fair" scheduler to run it too often in the detriment of other generators.
The intention was to run it every 2 hours but the measurement unit for 7200 was not specified.
Add more logs, including showing a generator name, in order to troubleshoot this kind of issues easier in the future.
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
- update cve tests
- update scrub tests
- update tests for parsing storage and loading into meta DB
- update controller tests
Signed-off-by: Andrei Aaron <aaaron@luxoft.com>
instead of reading entire files before calculating their digests
stream them by using their Reader method.
Signed-off-by: Petu Eusebiu <peusebiu@cisco.com>
* fix(scheduler): data race when pushing new tasks
the problem here is that scheduler can be closed in two ways:
- canceling the context given as argument to scheduler.RunScheduler()
- running scheduler.Shutdown()
because of this shutdown can trigger a data race between calling scheduler.inShutdown()
and actually pushing tasks into the pool workers
solved that by keeping a quit channel and listening on both quit channel and ctx.Done()
and closing the worker chan and scheduler afterwards.
Signed-off-by: Petu Eusebiu <peusebiu@cisco.com>
* refactor(scheduler): refactor into a single shutdown
before this we could stop scheduler either by closing the context
provided to RunScheduler(ctx) or by running Shutdown().
simplify things by getting rid of the external context in RunScheduler().
keep an internal context in the scheduler itself and pass it down to all tasks.
Signed-off-by: Petu Eusebiu <peusebiu@cisco.com>
---------
Signed-off-by: Petu Eusebiu <peusebiu@cisco.com>
wait for workers to finish before exiting
should fix tests reporting they couldn't remove rootDir because it's being
written by tasks
Signed-off-by: Petu Eusebiu <peusebiu@cisco.com>
Andrei Aaron