zhangyue19921010 opened a new issue, #11283:
URL: https://github.com/apache/paimon/issues/11283

   ### Search before asking
   
   - [x] I searched in the issues and found no report of this duration 
calculation across daylight-saving changes.
   
   ### Paimon version
   
   Current master, `3a8e01f35d8fc1391f86286001a98340d65d7f80` (2.2-SNAPSHOT).
   
   ### Compute Engine
   
   Java table API; consumer expiration runs during table commits.
   
   ### Minimal reproduce step
   
   Use a table with `consumer.expiration-time=1 h`, 
`snapshot.num-retained.min=1`, and `snapshot.num-retained.max=1`. Record a 
consumer pointing to the first snapshot, then commit another snapshot with the 
writer in `America/New_York`.
   
   - Spring forward: at `2026-03-08T07:30:00Z` (03:30 local), a consumer last 
modified at `06:45Z` (01:45 local) is only 45 minutes old. The local cutoff 
becomes 02:30, so the active consumer is deleted early and its snapshot 
protection is removed.
   - Fall back: at `2026-11-01T06:30:00Z` (the second 01:30), a consumer last 
modified at `05:15Z` is 75 minutes old. The local cutoff is 00:30, so the stale 
consumer is retained.
   
   A persisted-table regression fixes the clock at these instants and sets the 
consumer file's real modification time. Both cases fail on unchanged master; 
ordinary-day, UTC and exact-boundary controls pass.
   
   ### What doesn't meet your expectations?
   
   The option promises a lifetime since the consumer file's last modification. 
Duration arithmetic should use elapsed time and compare against the stored 
modification timestamp, without a daylight-saving offset change adding or 
removing an hour.
   
   ### Anything else?
   
   The commit path can use an epoch-millisecond cutoff. Keep the existing 
`expire(LocalDateTime)` entrypoint and its semantics for callers that 
explicitly pass a local-time cutoff.
   
   ### Are you willing to submit a PR?
   
   - [x] I'm willing to submit a PR!
   


-- 
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