Part 1 includes
* New experimental feature to enable the new code
* New providers using RemoteCache only
* New test profile to run the tests with the experimental feature
New providers' implementation for:
* InfinispanConnectionProvider
* AuthenticationSessionProvider
* ClusterProvider
Closes#28140
Signed-off-by: Pedro Ruivo <pruivo@redhat.com>
For String fields that may be null, convert an empty string to null when
reading from Protostream
Fixes#30511
Signed-off-by: Pedro Ruivo <pruivo@redhat.com>
The field lastIPFailure can be null and needs a proto factory to set it
to null when missing.
Closes#30485
Signed-off-by: Pedro Ruivo <pruivo@redhat.com>
- Add tests for crud operations on configurable required actions
- Add support exposing the required action configuration via RequiredActionContext
- Make configSaveError message reusable in other contexts
- Introduced admin-ui specific endpoint for retrieving required actions with config metadata
Fixes#28400
Signed-off-by: Thomas Darimont <thomas.darimont@googlemail.com>
Co-authored-by: Erik Jan de Wit <erikjan.dewit@gmail.com>
The cache replace logic depends on it, as values returned by reference from a local cache must never be modified on those critical fields directly.
Closes#28906
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
Signed-off-by: Michal Hajas <mhajas@redhat.com>
Co-authored-by: Michal Hajas <mhajas@redhat.com>
Closes#28303
Signed-off-by: Pedro Ruivo <pruivo@redhat.com>
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
Co-authored-by: Alexander Schwartz <aschwart@redhat.com>
* Execute persistent sessions tests in CI and fix deadlock in UserSessionConcurrencyTest
The deadlock was caused by a switch to reactive handling of embedded Infinispan updates introduced here #28862Closes#29235
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
All writes for the sessions are handled by a background thread which batches them.
Closes#28862
Wait for persistent-store to contain update
instead of cache which has the change immediately since it is in memory + introduce new model-test profile
Closes#29141
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
Signed-off-by: Michal Hajas <mhajas@redhat.com>
Co-authored-by: Michal Hajas <mhajas@redhat.com>
Defer updates of last session refreshes and batch them
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
Signed-off-by: Michal Hajas <mhajas@redhat.com>
Co-authored-by: Michal Hajas <mhajas@redhat.com>
+ since the runSerialized mechanism is currently on the best effort basis it is possible there are concurrent executions if T1 obtained a lock T2 removed the lock and T3 created a new lock before T1 called putIfAbsent therefore I added a debug log detecting this situation
Signed-off-by: Michal Hajas <mhajas@redhat.com>
Signed-off-by: vibrown <vibrown@redhat.com>
More updates
Signed-off-by: vibrown <vibrown@redhat.com>
Added client type logic from Marek's prototype
Signed-off-by: vibrown <vibrown@redhat.com>
updates
Signed-off-by: vibrown <vibrown@redhat.com>
updates
Signed-off-by: vibrown <vibrown@redhat.com>
updates
Signed-off-by: vibrown <vibrown@redhat.com>
Testing to see if skipRestart was cause of test failures in MR
Adding two feature toggles for new code paths to store online sessions in the existing offline sessions table. Separate the code which is due to be changed in the next iteration in new classes/providers which used instead of the old one.
Closes#27976
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
Signed-off-by: Michal Hajas <mhajas@redhat.com>
Co-authored-by: Michal Hajas <mhajas@redhat.com>
Closes#25676
Signed-off-by: vramik <vramik@redhat.com>
Signed-off-by: Pedro Ruivo <pruivo@redhat.com>
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
Co-authored-by: Pedro Ruivo <pruivo@redhat.com>
Co-authored-by: Alexander Schwartz <aschwart@redhat.com>
Closes#26941
Signed-off-by: Michal Hajas <mhajas@redhat.com>
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
Co-authored-by: Michal Hajas <mhajas@redhat.com>
Closes#26810
Co-authored-by: Thomas Darimont <thomas.darimont@googlemail.com>
Co-authored-by: Martin Kanis <mkanis@redhat.com>
Signed-off-by: Thomas Darimont <thomas.darimont@googlemail.com>
Signed-off-by: Martin Kanis <mkanis@redhat.com>
This PR introduces a String getTaskName() default method to
the ScheduledTask interface and adjusts call sites to use the
implementation derived task name where possible.
Previously, ScheduledTask names were passed around separately, which
lead to unhelpful debug messages.
We now give ScheduledTask implementations control over their task-name
which allows for more flexible naming.
Enlist call StoreSyncEvent.fire(...) to after transaction to ensure realm is present in database.
Ensure that Realm is already committed before updating sync via UserStorageSyncManager
Align Sync task name generation for cancellation to support SyncFederationTest
Only log a message if sync task was actually canceled.
Signed-off-by: Thomas Darimont <thomas.darimont@googlemail.com>
Closes#26460
Signed-off-by: Václav Muzikář <vmuzikar@redhat.com>
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
Co-authored-by: Alexander Schwartz <aschwart@redhat.com>
Closes#22922
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
Signed-off-by: Pedro Ruivo <pruivo@redhat.com>
Co-authored-by: Pedro Ruivo <pruivo@redhat.com>
Closes#25077
Signed-off-by: Michal Hajas <mhajas@redhat.com>
Signed-off-by: Alexander Schwartz <aschwart@redhat.com>
Co-authored-by: Pedro Ruivo <pruivo@redhat.com>
Co-authored-by: Alexander Schwartz <aschwart@redhat.com>
closes#22372
Co-authored-by: Erik Jan de Wit <erikjan.dewit@gmail.com>
Co-authored-by: Pedro Igor <pigor.craveiro@gmail.com>
Co-authored-by: Michal Hajas <mhajas@redhat.com>
The problem is again the wrap(...) function.
In case the user is not found, then null is
returned. This can happen when a federated user
is deleted on the federation side but Keycloak
is not informed about it. In that case, the
session is still present but no UserModel can
be created.
Without this patch the stream contains null
values. Some downstream users can not cope well
with that.
The adjustment of the function getUserSessionsCount(...)
is slightly more expensive in execution, but
returns the correct number.
Closes#22428
Co-authored-by: Martin Krüger <mkrueger@mkru.de>
