* fix(ssubscribe): properly resubscribe in case of shard failover
1) when RE failover happens, there is a disconnect
2) affected Client reconnects and tries to resubscribe all existing listeners
ISSUE #1: CROSSSLOT Error - client was doing ssubscribe ch1 ch2.. chN which, after the failover could result in CROSSSLOT ( naturally, becasuse now some slots could be owned by other shards )
FIX: send one ssubscribe command per channel instead of one ssubscribe for all channels
ISSUE #2: MOVED Error - some/all of the channels might be moved somewhere else
FIX: 1: Propagate the error to the Cluster. 2: Cluster rediscovers topology.
3: Cluster resubscribes all listeners of the failed client ( possibly some/all of those will end up in a different client after the rediscovery )
fixes: #2902
Certain command parameter enums, specifically ClientKillFilters for the CLIENT KILL command, were not publicly exported by the @redis/client package. The purpose of this change is to make the ClientKillFilters enum accessible to all package consumers, eliminating the need for users to rely on an unstable internal import path.
This matters because it provides a reliable and supported way for developers to use commands that require these enums, improving the package's usability and API stability.
fixes: #2805
Co-authored-by: Nikolay Karadzhov <nkaradzhov89@gmail.com>
* rename maint options according to the latest client options
* adjust env variables
cae repo uses RE_FAULT_INJECTOR_URL for fault injector
DATABASE_NAME is needed to choose from the many databases in cae
* fix connection cleanup test
* feat(errors): Add specialized timeout error types for maintenance scenarios
- Added `SocketTimeoutDuringMaintananceError`, a subclass of `TimeoutError`, to handle socket timeouts during maintenance.
- Added `CommandTimeoutDuringMaintenanceError`, another subclass of `TimeoutError`, to address command write timeouts during maintenance.
* feat(linked-list): Add EmptyAwareSinglyLinkedList and enhance DoublyLinkedList functionality
- Introduced `EmptyAwareSinglyLinkedList`, a subclass of `SinglyLinkedList` that emits an `empty` event when the list becomes empty due to `reset`, `shift`, or `remove` operations.
- Added `nodes()` iterator method to `DoublyLinkedList` for iterating over nodes directly.
- Enhanced unit tests for `DoublyLinkedList` and `SinglyLinkedList` to cover edge cases and new functionality.
- Added comprehensive tests for `EmptyAwareSinglyLinkedList` to validate `empty` event emission under various scenarios.
- Improved code formatting and consistency.
* refactor(commands-queue): Improve push notification handling
- Replaced `setInvalidateCallback` with a more flexible `addPushHandler` method, allowing multiple handlers for push notifications.
- Introduced the `PushHandler` type to standardize push notification processing.
- Refactored `RedisCommandsQueue` to use a `#pushHandlers` array, enabling dynamic and modular handling of push notifications.
- Updated `RedisClient` to leverage the new handler mechanism for `invalidate` push notifications, simplifying and decoupling logic.
* feat(commands-queue): Add method to wait for in-flight commands to complete
- Introduced `waitForInflightCommandsToComplete` method to asynchronously wait for all in-flight commands to finish processing.
- Utilized the `empty` event from `#waitingForReply` to signal when all commands have been completed.
* feat(commands-queue): Introduce maintenance mode support for commands-queue
- Added `#maintenanceCommandTimeout` and `setMaintenanceCommandTimeout` method to dynamically adjust command timeouts during maintenance
* refator(client): Extract socket event listener setup into helper method
* refactor(socket): Add maintenance mode support and dynamic timeout handling
- Added `#maintenanceTimeout` and `setMaintenanceTimeout` method to dynamically adjust socket timeouts during maintenance.
* feat(client): Add Redis Enterprise maintenance configuration options
- Added `maintPushNotifications` option to control how the client handles Redis Enterprise maintenance push notifications (`disabled`, `enabled`, `au
to`).
- Added `maintMovingEndpointType` option to specify the endpoint type for reconnecting during a MOVING notification (`auto`, `internal-ip`, `external-ip`,
etc.).
- Added `maintRelaxedCommandTimeout` option to define a relaxed timeout for commands during maintenance.
- Added `maintRelaxedSocketTimeout` option to define a relaxed timeout for the socket during maintenance.
- Enforced RESP3 requirement for maintenance-related features (`maintPushNotifications`).
* feat(client): Add socket helpers and pause mechanism
- Introduced `#paused` flag with corresponding `_pause` and `_unpause` methods to
temporarily halt writing commands to the socket during maintenance windows.
- Updated `#write` method to respect the `#paused` flag, preventing new commands from being written during maintenance.
- Added `_ejectSocket` method to safely detach from and return the current socket
- Added `_insertSocket` method to receive and start using a new socket
* feat(client): Add Redis Enterprise maintenance handling capabilities
- Introduced `EnterpriseMaintenanceManager` to manage Redis Enterprise maintenance events and push notifications.
- Integrated `EnterpriseMaintenanceManager` into `RedisClient` to handle maintenance push notifications and manage socket transitions.
- Implemented graceful handling of MOVING, MIGRATING, and FAILOVER push notifications, including socket replacement and timeout adjustments.
* test: add E2E test infrastructure for Redis maintenance scenarios
* test: add E2E tests for Redis Enterprise maintenance timeout handling (#3)
* test: add connection handoff test
---------
Co-authored-by: Pavel Pashov <pavel.pashov@redis.com>
Co-authored-by: Pavel Pashov <60297174+PavelPashov@users.noreply.github.com>
RedisClusterType was using some type restrictions to
hide some commands based on its properties. In reality,
all commands should be exposed to user ( as in v4 ).
fixes#3064
