ryanmeowy opened a new issue, #13146:
URL: https://github.com/apache/seatunnel/issues/13146

   ### Search before asking
   
   - [x] I had searched in the 
[issues](https://github.com/apache/seatunnel/issues?q=is%3Aissue+label%3A%22bug%22)
 and found no similar issues.
   
   Searched (2026-11-05) in both issues and pull requests for `min-pause`,
   `AbstractFlinkRuntimeEnvironment`, `setMinPauseBetweenCheckpoints`, 
`checkpoint.timeout min-pause`,
   `Flink checkpoint timeout` and `Flink checkpoint min pause`.
   
   Nothing reports this defect. The closest hits are unrelated:
   
   - #13097 — a docs PR that mentions `min-pause` in Zeta documentation.
   - #10524 — Paimon does not generate new snapshots on the Flink engine 
(connector issue).
   - #12789 / #12149 / #12339 — Zeta checkpoint scheduling work, engine side.
   - #13115 / #13081 — `[Fix][Core]` PRs that touch other files in
     `org.apache.seatunnel.core.starter.flink` (`SinkExecuteProcessor`, 
`FlinkCommandArgs`), not
     `AbstractFlinkRuntimeEnvironment`.
   
   ### What happened
   
   On the Flink engine, `env.min-pause` is silently dropped whenever 
`checkpoint.timeout` is also
   configured. The job starts normally, no warning is logged, and Flink keeps 
the default minimum
   pause of 0 ms, so the configured throttle is simply absent.
   
   `min-pause` is a documented, job-level `env` option
   (`docs/en/architecture/fault-tolerance/checkpoint-mechanism.md` shows it 
together with
   `checkpoint.interval` and `checkpoint.timeout` in the same `env` block), and 
it is accepted by the
   config validator as an optional env key, so a config that passes `--check` 
still loses the setting
   at runtime.
   
   Two code paths trigger it: the env key `checkpoint.timeout` and the legacy 
key
   `execution.checkpoint.timeout`.
   
   Workaround until this is fixed: the legacy key 
`execution.checkpoint.min-pause` (instead of
   `min-pause`) is applied on its own line and does take effect with or without 
a timeout.
   
   ### SeaTunnel Version
   
   dev (`a3743d7f52e9200a7733857ca813a0836bf66bba`)
   
   ### SeaTunnel Config
   
   ```conf
   env {
     parallelism = 1
     job.mode = "STREAMING"
     checkpoint.interval = 60000
     checkpoint.timeout  = 120000
     min-pause           = 10000
   }
   source {
     FakeSource {
       plugin_input = "fake"
       row.num = 1
       schema {
         fields {
           id = int
         }
       }
     }
   }
   sink {
     Console {
       plugin_input = "fake"
     }
   }
   ```
   
   ### Running Command
   
   ```shell
   ./bin/start-seatunnel-flink-15-connector-v2.sh --config ./min-pause.conf
   ```
   
   ### Error Exception
   
   There is no exception — that is the problem. The config is dropped silently 
and the job runs with
   a minimum pause of 0 ms. Observable in the Flink Web UI (Checkpoints → *Min 
Pause Between
   Checkpoints* shows `0 ms` although `min-pause = 10000` is configured), and 
in the checkpoint
   history: nothing keeps two checkpoints apart, so a slow checkpoint is 
followed immediately by the
   next trigger instead of waiting for the configured 10 s gap.
   
   Measured on the same build, toggling only the `checkpoint.timeout` line in 
the config above and
   reading 
`getStreamExecutionEnvironment().getCheckpointConfig().getMinPauseBetweenCheckpoints()`:
   
   ```
   config = interval + timeout + min-pause   ->  minPauseBetweenCheckpoints = 0 
     (expected 10000)
   config = interval + min-pause             ->  minPauseBetweenCheckpoints = 
10000
   ```
   
   The same happens with the legacy key (`env { execution.checkpoint.timeout = 
120000, min-pause = 10000 }`).
   
   ### Zeta or Flink or Spark Version
   
   Flink 1.15.3 (`seatunnel-flink-15-starter`). The affected class is shared by 
the
   `seatunnel-flink-13-starter`, `seatunnel-flink-15-starter` and 
`seatunnel-flink-20-starter`
   modules, so all three are affected.
   
   ### Java or Scala Version
   
   JDK 21 (`21.0.12.1`) used for the reproduction; the project targets Java 11.
   
   ### Are you willing to submit PR?
   
   - [x] Yes I am willing to submit a PR!
   
   ### Code of Conduct
   
   - [x] I agree to follow this project's [Code of 
Conduct](https://www.apache.org/foundation/policies/conduct)
   


-- 
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