[ 
https://issues.apache.org/jira/browse/IGNITE-28700?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Vyacheslav Koptilin updated IGNITE-28700:
-----------------------------------------
    Labels: ignite-3  (was: )

> CDC / Change Data Capture: no streaming change feed in Ignite 3.x
> -----------------------------------------------------------------
>
>                 Key: IGNITE-28700
>                 URL: https://issues.apache.org/jira/browse/IGNITE-28700
>             Project: Ignite
>          Issue Type: Bug
>          Components: replication ai3
>         Environment: Apache Ignite:3.1.0 (migrated from 2.17.0)
> JDK:Java 21 (LTS)
>  
> Topology:Multi-node clusters (2–3 nodes SharedNothing, both same-host and 
> cross-host
> deployments). Distribution zones with `PARTITIONS=25`, `REPLICAS` tuned per 
> zone.
>  
> Workload:CEP engine — high write rate on entity / event tables (rules generate
>  thousands of writes per second), all of which must be observable by 
> downstream consumers
>  (rule re-evaluation across nodes, audit, optional Kafka mirroring).
> Persistence: `aipersist` regions with checkpointing; `raft.fsync=true` for 
> SharedNothing.
>            Reporter: Kratika Agrawal
>            Priority: Major
>              Labels: ignite-3
>
> Ignite 2.x supported CDC via `CacheEntryListener` (JCache) and the cache 
> change-event SPI,
> enabling streaming of cache mutations to Kafka, Pulsar, or other external 
> systems for audit,
> search-index hydration, or downstream materialization.
> Ignite 3.1.0 removed both `CacheEntryListener` and the underlying 
> continuous-query / change-event
> SPI. There is no streaming API (push or pull) for "tell me everything that 
> changed in this
> table since timestamp/offset X".
> Ask:
> Document the intended 3.x pattern** for downstream replication / 
> event-sourcing. Today's
>    options (compute job broadcasts? application-side changelog tables? raft 
> log tailing?) are
>    not specified.
> 3. **Accept this as a regression** — 2.x users with downstream consumers 
> (Kafka, Elasticsearch,
>    audit) cannot migrate without rewriting their integration layer.



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

Reply via email to