HDDS-15562. EventNotification: persistence strategy for consumer checkpoint (via distributed RocksDB update)#10689
Open
gardenia wants to merge 1 commit into
Conversation
gardenia
changed the base branch from
master
to
HDDS-13513_Event_Notification_FeatureBranch
Author
|
NOTE: this is a rework of the original implementation which used Ozone filesystem bucket property metadata to store the properties (PR and review comments from @ChenSammi can be found here #10515). This approach of using a distributed write to the RocksDB metadata table for persisting the checkpoint was suggested by @ChenSammi in review comments you can find on the above PR. |
Persist the checkpoint within the OM RocksDB metadata table using a distributed Ratis update message so that another OM can pick up if the leader changes. Technical Solution: * Update the RocksDB metadata table (OMSetEventNotificationCheckpointRequest) and distribute this to other OMs via Ratis (SetEventNotificationCheckpointRequest) Checkpoint Throttling: * We update our local progress in-memory on every transaction batch, but throttle the persistent Ratis writes (configurable via OZONE_OM_PLUGIN_CHECKPOINT_SAVE_INTERVAL, defaulting to every 100 transactions). * The intent is to prevent saturating the Ratis log or causing write storms on RocksDB while maintain our at-least-once delivery guarantee. * If a Ratis write fails, throttling is bypassed to ensure the next update cycle retries immediately. Error Handling: * Both loading and saving failures are non-blocking for background polling progress * On load failures, we log a warning and fallback to starting from transaction 0/null. * On save failures, we log a warning but always advance progress in-memory to ensure progress but conform to at-least-once event delivery. Detailed Changes: - Protobuf: - Add SetEventNotificationCheckpoint type and request/response payloads to OmClientProtocol.proto. - OmUtils & OzoneManagerRatisUtils: - Wire SetEventNotificationCheckpoint as a write request. - Route SetEventNotificationCheckpoint to its corresponding OMRequest class. - Requests & Responses: - Create OMSetEventNotificationCheckpointRequest to write checkpoint values to metaTable. - Create OMSetEventNotificationCheckpointResponse to persist values in metaTable via addToDBBatch. - Strategy: - Expose NotificationCheckpointStrategy interface with load, save, and reset. - Implement OzoneDbCheckpointStrategy to query and save checkpoints atomically via the new OM requests. - Added reset() call to durably clear checkpoints during self-healing. - Plugin Context & Poller: - Wire the checkpoint strategy through OMEventListenerPluginContext and OMEventListenerPluginManager. - Integrate strategy into the poller and publisher. Both loading and saving failures are non-blocking, falling back to starting from 0/null on load failures and continuing in-memory progress on save failures.
gardenia
force-pushed
the
HDDS-15562-DB-Checkpoint
branch
from
8936a16 to
2ed2e10
Compare
Contributor
There was a problem hiding this comment.
Pull request overview
This PR adds a durable, HA-consistent persistence mechanism for the OM event-notification consumer checkpoint by storing it in the OM RocksDB metaTable and replicating updates via a new Ratis-distributed OMRequest. This enables another OM to continue from the last persisted checkpoint after leader changes, while throttling persistent writes to avoid Ratis/RocksDB write storms.
Changes:
- Introduces a
NotificationCheckpointStrategyabstraction and an OM-backed implementation (OzoneDbCheckpointStrategy) that persists checkpoints via a newSetEventNotificationCheckpointOMRequest. - Wires checkpoint strategy into the plugin context and updates the Kafka publisher/poller seek position to load/save checkpoints non-blockingly.
- Extends the OM protocol (proto + request factory + routing helpers) and adds unit tests covering request handling and strategy throttling/failure behavior.
Reviewed changes
Copilot reviewed 18 out of 18 changed files in this pull request and generated 3 comments.
Show a summary per file
| File | Description |
|---|---|
| hadoop-ozone/ozone-manager/src/test/java/org/apache/hadoop/ozone/om/request/eventlistener/TestOMSetEventNotificationCheckpointRequest.java | Adds UT for the new OM request caching/metaTable behavior. |
| hadoop-ozone/ozone-manager/src/test/java/org/apache/hadoop/ozone/om/eventlistener/TestOzoneDbCheckpointStrategy.java | Adds UTs for load/save throttling, failure recovery, and reset behavior. |
| hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/response/eventlistener/package-info.java | New response package documentation. |
| hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/response/eventlistener/OMSetEventNotificationCheckpointResponse.java | New response that batches metaTable checkpoint writes. |
| hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/request/eventlistener/package-info.java | New request package documentation. |
| hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/request/eventlistener/OMSetEventNotificationCheckpointRequest.java | New request that updates the metaTable cache for checkpoint persistence. |
| hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/ratis/utils/OzoneManagerRatisUtils.java | Registers the new request type in the request factory. |
| hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/eventlistener/OzoneDbCheckpointStrategy.java | Implements checkpoint persistence via Ratis-submitted OMRequest with throttling. |
| hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/eventlistener/OMEventListenerPluginManager.java | Creates and injects the checkpoint strategy into the plugin context. |
| hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/eventlistener/OMEventListenerPluginContextImpl.java | Exposes checkpoint strategy via the plugin context implementation. |
| hadoop-ozone/ozone-manager-plugins/src/test/java/org/apache/hadoop/ozone/om/eventlistener/TestOMEventListenerLedgerPollerRecovery.java | Adds UTs ensuring load/save failures don’t block poller progress. |
| hadoop-ozone/ozone-manager-plugins/src/main/java/org/apache/hadoop/ozone/om/eventlistener/OMEventListenerLedgerPollerSeekPosition.java | Loads/saves checkpoint via strategy with non-blocking failure handling and reset. |
| hadoop-ozone/ozone-manager-plugins/src/main/java/org/apache/hadoop/ozone/om/eventlistener/OMEventListenerKafkaPublisher.java | Uses plugin-provided checkpoint strategy for seek position initialization. |
| hadoop-ozone/interface-client/src/main/proto/OmClientProtocol.proto | Adds new OMRequest/OMResponse types and messages for checkpoint updates. |
| hadoop-ozone/common/src/main/java/org/apache/hadoop/ozone/OmUtils.java | Classifies the new request as a write and leader-routed request. |
| hadoop-ozone/common/src/main/java/org/apache/hadoop/ozone/om/eventlistener/OMEventListenerPluginContext.java | Adds an API to obtain the checkpoint strategy from plugin context. |
| hadoop-ozone/common/src/main/java/org/apache/hadoop/ozone/om/eventlistener/NotificationCheckpointStrategy.java | Adds new strategy interface for checkpoint persistence. |
| hadoop-hdds/common/src/main/java/org/apache/hadoop/ozone/OzoneConsts.java | Adds a dedicated metaTable key prefix for event-notification checkpoints. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+54
to
+62
| final SetEventNotificationCheckpointRequest request = | ||
| getOmRequest().getSetEventNotificationCheckpointRequest(); | ||
|
|
||
| final String checkpointKey = request.getCheckpointKey(); | ||
| final String checkpointValue = request.getCheckpointValue(); | ||
|
|
||
| // Store in metaTable using CacheKey with our specific checkpoint prefix | ||
| final String dbKey = OzoneConsts.EVENT_NOTIFICATION_CHECKPOINT_PREFIX + checkpointKey; | ||
|
|
Comment on lines
34
to
+37
| List<OmCompletedRequestInfo> listCompletedRequestInfo(Long startKey, int maxResults) throws IOException; | ||
|
|
||
| NotificationCheckpointStrategy getNotificationCheckpointStrategy(); | ||
|
|
Comment on lines
+133
to
+135
| submitRequest(omRequest); | ||
| LOG.info("Persisted {} = {} directly as metadata inside metaTable under key {}", | ||
| checkpointName, val, dbKey); |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Please describe your PR in detail:
Persist the checkpoint within the OM RocksDB metadata table using a distributed Ratis update message so that another OM can pick up if the leader changes.
Technical Solution:
Checkpoint Throttling:
Error Handling:
What is the link to the Apache JIRA
https://issues.apache.org/jira/browse/HDDS-15562
How was this patch tested?
Unit tests