[jira] [Commented] (KAFKA-14374) Kafka streams losing messages in State Store during first launch of app

["Matthias J. Sax (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/KAFKA-14374?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17635667#comment-17635667 ] 

Matthias J. Sax commented on KAFKA-14374:
-----------------------------------------

Thanks for creating a ticket. I am not sure right now if I can follow all details. Couple of questions:
{quote}we should expect that #table2=#repartition-topic=#state-store
{quote}
What do you mean by this? Do you expect that an input record from partition 0 (table2 topic) will go to repartition-topic partition 0? But default, this won't be guaranteed, because the default hash-partitioning may compute a different partition for an integer vs a string.
{quote}which actually is not verified.
{quote}
Not sure if I can follow.
{quote}What we end up with is the following #table2=#repartition-topic, but  #repartition-topic>#state-store.
{quote}
Not sure what this means. Can you elaborate?
{quote}Please note that there is no insertion in `table2` as we paused the connector to verify the cardinality.
{quote}
How do you do the verification?
{quote}So by creating the internal topics before, we avoid the rebalance and we end up by #table2=#repartition-topic=#state-store.
{quote}
There will _always_ be a rebalance on first startup...

Your understanding of the rebalance protocol seems not to be correct. The group coordinator only "moderates" a rebalance, but it does not compute any assignment. One of the consumer inside the group is elected as "group leader" and this group leader computes the partition assignment. During assignment computation, the group leader also checks if internal topic exists, and creates them if necessary. (This topic creations happens before the assignment is computed.)
{quote}which triggers the delete of those messages without flushing them to changelog topic.
{quote}
What do you mean by "triggers the delete"? Also, why do think that changelog writes are not flushed?

> Kafka streams losing messages in State Store during first launch of app
> -----------------------------------------------------------------------
>
>                 Key: KAFKA-14374
>                 URL: https://issues.apache.org/jira/browse/KAFKA-14374
>             Project: Kafka
>          Issue Type: Bug
>          Components: streams
>    Affects Versions: 2.8.0, 3.2.0
>            Reporter: Youcef Sebiat
>            Priority: Major
>         Attachments: Screenshot 2022-11-09 at 14.56.00.png
>
>
> We have been using Kafka Streams to implement a CDC based app. Attached is the sub topology of interest.
> `table2` topic is created by Debezium who is connected to a SQL DB. It contains 26K lines. We take `table2` and create a key which is only a conversion of the key from `string` to `int`. This means that we should expect that #table2=#repartition-topic=#state-store; which actually is not verified. What we end up with is the following #table2=#repartition-topic, but  #repartition-topic>#state-store. We actually lose messages and thus corrupt the state store, which makes the app live in incorrect state. (Please note that there is no insertion in `table2` as we paused the connector to verify the cardinality.)
> The above happens only during the first launch, i.e. the app has never been launched before, so internal topics do not exist yet. Restarts of pre-existing apps do not yield any problems.
> We have:
> 1. Broker on Kafka 3.2.
> 2. Client app on 2.8|3.2 (we tried both and we faced the same issue).
> 2. All parameters are by default except `CACHE_MAX_BYTES_BUFFERING_CONFIG` set to `0` and `NUM_STREAM_THREADS_CONFIG` set to `>1`.
>  
> *What actually worked*
> 1. Use a monothread at first launch: using one thread solves the problem. The #table2=#repartition-topic=#state-store is verified.
> 2. Pre-creating kafka internal topics: we noticed that whenever there is rebalance during the first launch of Kafka Streams app, the state stores ended up missing values. This also happens when you launch multiple pods in K8s for example. When we read through the logs, we noticed that there is a rebalance that is triggered when we first launch the app. This comes from the fact that the internal topics get created and assigned, thus the rebalance. So by creating the internal topics before, we avoid the rebalance and we end up by #table2=#repartition-topic=#state-store.
> *What we noticed from the logs*
> On multi-thread mode, we noticed that it is the partitions that are assigned to the thread chosen by the Coordinator to be the Leader of consumers that suffer the data loss. What we think is happening is the following:
> 1. Consumers threads are launched and inform the coordinator.
> 2. Coordinator assign topics and choses the Leader among the threads.
> 3. App create internal topics.
> 4. Consumers/producers process data. Specifically the Consumer leader consumes from the repartition topic, which triggers the delete of those messages without flushing them to changelog topic.
> 5. Leader notified of new assignment with internal topics. Triggers rebalance.
> 6. Leader pauses partitions. 
> 7. Rebalance finished. The leader resumes partitions.
> 8. Leader fetches the oldest offset of repartition partitions he got assigned. He will not start from zero, but instead from where he got interrupted in 4. The chunk of early messages are thus lost.
> Please note, that on mono-thread mode, there is no data loss which is weird since the leader is actually the unique thread. 
> So my questions are:
> 1. Are we understanding wrongly what's happening?
> 2. What can be the origin of this problem?



--
This message was sent by Atlassian Jira
(v8.20.10#820010)