jolshan commented on code in PR #14774:
URL: https://github.com/apache/kafka/pull/14774#discussion_r1420797233
##########
core/src/main/scala/kafka/server/ReplicaManager.scala:
##########
@@ -976,6 +975,231 @@ class ReplicaManager(val config: KafkaConfig,
}
}
+ /**
+ * Append messages to offsets topic, and wait for them to be replicated to
other replicas;
+ * the callback function will be triggered either when timeout or the
required acks are satisfied;
+ * if the callback function itself is already synchronized on some object
then pass this object to avoid deadlock.
+ *
+ * Noted that all pending delayed check operations are stored in a queue.
All callers to ReplicaManager.appendRecords()
+ * are expected to call ActionQueue.tryCompleteActions for all affected
partitions, without holding any conflicting
+ * locks.
+ *
+ * @param timeout maximum time we will wait to append
before returning
+ * @param requiredAcks number of replicas who must
acknowledge the append before sending the response
+ * @param internalTopicsAllowed boolean indicating whether internal
topics can be appended to
+ * @param origin source of the append request (ie,
client, replication, coordinator)
+ * @param entriesPerPartition the records per partition to be
appended
+ * @param responseCallback callback for sending the response
+ * @param delayedProduceLock lock for the delayed actions
+ * @param recordValidationStatsCallback callback for updating stats on
record conversions
+ * @param requestLocal container for the stateful instances
scoped to this request -- this must correspond to the
+ * thread calling this method
+ * @param actionQueue the action queue to use.
ReplicaManager#defaultActionQueue is used by default.
+ * @param verificationGuards the mapping from topic partition to
verification guards if transaction verification is used
+ * @param preAppendErrors the mapping from topic partition to
LogAppendResult for errors that occurred before appending
+ */
+ def appendForGroup(timeout: Long,
Review Comment:
> this code must not return until the local write is done
I'm not sure I understand this point. This method doesn't return until the
local write is done. Are you suggesting in the future appendToLocalLog would be
async?
> call maybeStartTransactionVerificationForPartition with a callback that
would call this method
This will change when I do the refactor since this will become appendRecords
where we only call maybeStartTransactionVerificationForPartition(s) if the
append requires transaction verification. I can add the comment now, but it
will be changed in the refactor
(https://issues.apache.org/jira/browse/KAFKA-15987)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
