This repository has been archived by the owner on Apr 1, 2024. It is now read-only.
forked from apache/pulsar
-
Notifications
You must be signed in to change notification settings - Fork 26
Conversation
This file contains 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
…apache#20568) - Since `cnx.address + consumerId` is the identifier of one consumer; add `consumer-id` into the log when doing subscribe. - add a test to confirm that even if the error occurs when sending messages to the client, the consumption is still OK. - print debug log if ack-command was discarded due to `ConsumerFuture is not complete.` - print debug log if sending a message to the client is failed. (cherry picked from commit a41ac49)
…luster (apache#20767) (cherry picked from commit 465fac5)
…pache#20819) Motivation: If the producer name is generated by the Broker, the producer will update the variable `producerName` after connecting, but not update the same variable of the batch message container. Modifications: fix bug (cherry picked from commit aba50f2)
(cherry picked from commit 5b32220)
…egistered if there has no message was sent (apache#20888) Motivation: In the replication scenario, we want to produce messages on the native cluster and consume messages on the remote cluster, the producer and consumer both use a same schema, but the consumer cannot be registered if there has no messages in the topic yet.The root cause is that for the remote cluster, there is a producer who has been registered with `AUTO_PRODUCE_BYTES` schema, so there is no schema to check the compatibility. Modifications: If there is no schema and only the replicator producer was registered, skip the compatibility check. (cherry picked from commit 9be0b52)
- The task `trim ledgers` runs in the thread `BkMainThreadPool.choose(ledgerName)` - The task `write entries to BK` runs in the thread `BkMainThreadPool.choose(ledgerId)` So the two tasks above may run concurrently/ The task `trim ledgers` work as the flow below: - find the ledgers which are no longer to read, the result is `{Ledgers before the slowest read}`. - check if the `{Ledgers before the slowest read}` is out of retention policy, the result is `{Ledgers to be deleted}`. - if the create time of the ledger is lower than the earliest retention time, mark it should be deleted - if after deleting this ledger, the rest ledgers are still larger than the retention size, mark it should be deleted - delete the`{Ledgers to be deleted}` **(Highlight)** There is a scenario that causes the task `trim ledgers` did discontinuous ledger deletion, resulting consume messages discontinuous: - context: - ledgers: `[{id=1, size=100}, {id=2,size=100}]` - retention size: 150 - no cursor there - Check `ledger 1`, skip by retention check `(200 - 100) < 150` - One in-flight writing is finished, the `calculateTotalSizeWrited()` would return `300` now. - Check `ledger 2`, retention check `(300 - 100) > 150`, mark the ledger-2 should be deleted. - Delete the `ledger 2`. - Create a new consumer. It will receive messages from `[ledger-1, ledegr-3]`, but the `ledger-2` will be skipped. Once the retention constraint has been met, break the loop. (cherry picked from commit 782e91f)
The C++ and Python clients are not maintained in the main repo now.
(cherry picked from commit 3ab420c)
(cherry picked from commit 3ab420c)
Fixes: apache#20997 Update the expired certs to get tests passing. * Update all certs. See `README.md` in files for detailed steps. This change is covered by tests. - [x] `doc-not-needed` (cherry picked from commit d6734b7)
…r. (apache#20880) Main Issue: apache#20851 ### Motivation When the Proto version does not allow us to send TcClientConnectRequest to the broker, we should add a log to debug it. ### Modifications Add a waining log.
…edup and set namespace disable deduplication. (apache#20905)
…pache#21035) Motivation: After [PIP-118: reconnect broker when ZooKeeper session expires](apache#13341), the Broker will not shut down after losing the connection of the local metadata store in the default configuration. However, before the ZK client is reconnected, the events of BK online and offline are lost, resulting in incorrect BK info in the memory. You can reproduce the issue by the test `BkEnsemblesChaosTest. testBookieInfoIsCorrectEvenIfLostNotificationDueToZKClientReconnect`(90% probability of reproduce of the issue, run it again if the issue does not occur) Modifications: Refresh BK info in memory after the ZK client is reconnected. (cherry picked from commit db20035)
…he (apache#20763)" This reverts commit 1a7accb.
…ageId read reaches lastReadId (apache#20988) (cherry picked from commit 9e2195c)
…letion of compaction (apache#21067) (cherry picked from commit bb9c9b4)
…ing critical metadata operation paths (apache#18518) (cherry picked from commit 492a9c3)
…ata when doing lookup (apache#21063) Motivation: If we set `allowAutoTopicCreationType` to `PARTITIONED`, the flow of the create topic progress is like the below: 1. `Client-side`: Lookup topic to get partitioned topic metadata to create a producer. 1. `Broker-side`: Create partitioned topic metadata. 1. `Broker-side`: response `{"partitions":3}`. 1. `Client-side`: Create separate connections for each partition of the topic. 1. `Broker-side`: Receive 3 connect requests and create 3 partition-topics. In the `step 2` above, the flow of the progress is like the below: 1. Check the policy of topic auto-creation( the policy is `{allowAutoTopicCreationType=PARTITIONED, defaultNumPartitions=3}` ) 1. Check the partitioned topic metadata already exists. 1. Try to create the partitioned topic metadata if it does not exist. 1. If created failed by the partitioned topic metadata already exists( maybe another broker is also creating now), read partitioned topic metadata from the metadata store and respond to the client. There is a race condition that makes the client get non-partitioned metadata of the topic: | time | `broker-1` | `broker-2` | | --- | --- | --- | | 1 | get policy: `PARTITIONED, 3` | get policy: `PARTITIONED, 3` | | 2 | check the partitioned topic metadata already exists | Check the partitioned topic metadata already exists | | 3 | Partitioned topic metadata does not exist, the metadata cache will cache an empty optional for the path | Partitioned topic metadata does not exist, the metadata cache will cache an empty optional for the path | | 4 | | succeed create the partitioned topic metadata | | 5 | Receive a ZK node changed event to invalidate the cache of the partitioned topic metadata | | 6 | Creating the metadata failed due to it already exists | | 7 | Read the partitioned topic metadata again | If `step-5` is executed later than `step-7`, `broker-1` will get an empty optional from the cache of the partitioned topic metadata and respond non-partitioned metadata to the client. **What thing would make the `step-5` is executed later than `step-7`?** Provide a scenario: Such as the issue that the PR apache#20303 fixed, it makes `zk operation` and `zk node changed notifications` executed in different threads: `main-thread of ZK client` and `metadata store thread`. Therefore, the mechanism of the lookup partitioned topic metadata is fragile and we need to optimize it. Modifications: Before reading the partitioned topic metadata again, refresh the cache first. (cherry picked from commit d099ac4)
…ache#20948) Make the chunk message function work properly when deduplication is enabled. For example: ```markdown Chunk-1 sequence ID: 0, chunk ID: 0, total chunk: 2 Chunk-2 sequence ID: 0, chunk ID: 1 Chunk-3 sequence ID: 1, chunk ID: 0 total chunk: 3 Chunk-4 sequence ID: 1, chunk ID: 1 Chunk-5 sequence ID: 1, chunk ID: 1 Chunk-6 sequence ID: 1, chunk ID: 2 ``` Only store check and store the sequence ID of Chunk-2 and Chunk-6. **Add a property in the publishContext to determine whether this chunk is the last chunk when persistent completely.** ```java publishContext.setProperty(IS_LAST_CHUNK, Boolean.FALSE); ``` For example: ```markdown Chunk-1 sequence ID: 0, chunk ID: 0, msgID: 1:1 Chunk-2 sequence ID: 0, chunk ID: 1, msgID: 1:2 Chunk-3 sequence ID: 0, chunk ID: 2, msgID: 1:3 Chunk-4 sequence ID: 0, chunk ID: 1, msgID: 1:4 Chunk-5 sequence ID: 0, chunk ID: 2, msgID: 1:5 Chunk-6 sequence ID: 0, chunk ID: 3, msgID: 1:6 ``` We should filter and ack chunk-4 and chunk-5. (cherry picked from commit b0b13bc)
…che#21070) Current, when the producer resend the chunked message like this: - M1: UUID: 0, ChunkID: 0 - M2: UUID: 0, ChunkID: 0 // Resend the first chunk - M3: UUID: 0, ChunkID: 1 When the consumer received the M2, it will find that it's already tracking the UUID:0 chunked messages, and will then discard the message M1 and M2. This will lead to unable to consume the whole chunked message even though it's already persisted in the Pulsar topic. Here is the code logic: https://github.com/apache/pulsar/blob/44a055b8a55078bcf93f4904991598541aa6c1ee/pulsar-client/src/main/java/org/apache/pulsar/client/impl/ConsumerImpl.java#L1436-L1482 The bug can be easily reproduced using the testcase `testResendChunkMessages` introduced by this PR. - When receiving the new duplicated first chunk of a chunked message, the consumer discard the current chunked message context and create a new context to track the following messages. For the case mentioned in Motivation, the M1 will be released and the consumer will assemble M2 and M3 as the chunked message. (cherry picked from commit eb2e3a2)
## Motivation Handle ack hole case: For example: ```markdown Chunk-1 sequence ID: 0, chunk ID: 0, msgID: 1:1 Chunk-2 sequence ID: 0, chunk ID: 1, msgID: 1:2 Chunk-3 sequence ID: 0, chunk ID: 0, msgID: 1:3 Chunk-4 sequence ID: 0, chunk ID: 1, msgID: 1:4 Chunk-5 sequence ID: 0, chunk ID: 2, msgID: 1:5 ``` Consumer ack chunk message via ChunkMessageIdImpl that consists of all the chunks in this chunk message(Chunk-3, Chunk-4, Chunk-5). The Chunk-1 and Chunk-2 are not included in the ChunkMessageIdImpl, so we should process it here. ## Modification Ack chunk-1 and chunk-2. (cherry picked from commit 59a8e72)
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
This is a PR created by snbot to trigger the check suite in each repository.