DanielLeens commented on issue #11013:
URL: https://github.com/apache/seatunnel/issues/11013#issuecomment-4640972087

   That additional observation is useful.
   
   I checked the current reader flow more closely, and I do not see an 
intentional handoff that marks the snapshot as finished early in the normal 
path. In `INITIAL` mode, the reader only flips `snapshotCompleted` after 
`snapshotEvents(...)` returns, and `resolvedTs` is only advanced after the 
snapshot scan reaches the end of the current split.
   
   So the `~91,746` ceiling you observed is more likely a symptom of the same 
underlying reader bug family rather than a separate "expected switch to 
streaming" path. In other words, this still looks like either:
   1. the snapshot phase exited prematurely for some split, or
   2. the reader reached streaming with an already incomplete materialization 
state and then continued advancing `resolvedTs`.
   
   The most useful next evidence, if you still have it, would be:
   1. the startup log lines around `[splitId] Snapshot events start.` and the 
first `Capture streaming event from resolvedTs` for the same split
   2. how many source splits were assigned for that table in the job
   
   That will let us distinguish "snapshot never fully drained" from "streaming 
advanced after a partial snapshot" much more cleanly. Either way, this still 
looks like a real TiDB-CDC reader bug worth keeping open.


-- 
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]

Reply via email to