DanielCarter-stack commented on issue #10481:
URL: https://github.com/apache/seatunnel/issues/10481#issuecomment-3882015830
<!-- code-pr-reviewer -->
Thanks for the detailed report. This is confirmed as a design limitation.
**Root Cause:**
When `exactly_once` is enabled, the SQL Server CDC connector sets the
snapshot queue size to `Integer.MAX_VALUE` to buffer all snapshot data in
memory, causing OOM for large tables.
**Evidence:**
-
`connector/cdc/sqlserver-cdc/src/main/java/org/apache/seatunnel/connectors/seatunnel/cdc/sqlserver/source/reader/fetch/SqlServerSourceFetchTaskContext.java:121-124`
- `docs/en/connectors/source/SqlServer-CDC.md:20`
**Workarounds:**
1. **Disable exactly-once** (immediate fix)
```hocon
SqlServer-CDC {
exactly_once = false # uses configured queue size
}
```
2. **Use incremental mode** to skip initial snapshot
```hocon
startup.mode = "latest"
```
3. **Reduce snapshot fetch size**
```hocon
snapshot.fetch.size = 512
snapshot.split.size = 4096
```
4. **Split the 600+ tables into multiple jobs**
Could you share your current `exactly_once` configuration, JVM heap size per
worker, and whether at-least-once semantics would be acceptable?
--
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]
