You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@seatunnel.apache.org by "beta-zhou (via GitHub)" <gi...@apache.org> on 2026/08/26 01:19:15 UTC
[I] seatunel抽取Oracle数据库过程中,SCN后往回抽取,导致seatunel同步进程失败 [seatunnel]
beta-zhou opened a new issue, #11965:
URL: https://github.com/apache/seatunnel/issues/11965
背景说明:
1.数据流,以下是通过oracle OGG同步工具抽取生产系统的数据到镜像库,seatunel抽取镜像库数据同步到doris。
oracle生产环境 - OGG - oracle镜像库 - seatunel - doris数据库
2.现象:
由于OGG同步进程异常失败了,重建OGG同步进程(重建OGG抽取进程,重新初始化数据,重建OGG应用进程),OGG同步进程修复后,发现seatunel进程失败了,一直抽取步到镜像库的数据。
3.seatunel日志:(Caused by: io.debezium.DebeziumException: Online REDO LOG files or archive log files do not contain the offset scn 2509495408. Please perform a new snapshot.)
具体日志记录
2026-08-26 03:06:50,127 ERROR [o.a.s.e.s.d.p.PhysicalVertex ] [hz.main.generic-operation.thread-23] - Job (1119747415390289925), Pipeline: [(1/1)], task: [pipeline-1 [Source[0]-Oracle-CDC]-SourceTask (1/1)], taskGroupLocation: [TaskGroupLocation{jobId=1119747415390289925, pipelineId=1, taskGroupId=2}] end with state FAILED and Exception: java.lang.RuntimeException: One or more fetchers have encountered exception
at org.apache.seatunnel.connectors.seatunnel.common.source.reader.fetcher.SplitFetcherManager.checkErrors(SplitFetcherManager.java:147)
at org.apache.seatunnel.connectors.seatunnel.common.source.reader.SourceReaderBase.getNextFetch(SourceReaderBase.java:167)
at org.apache.seatunnel.connectors.seatunnel.common.source.reader.SourceReaderBase.pollNext(SourceReaderBase.java:93)
at org.apache.seatunnel.connectors.cdc.base.source.reader.IncrementalSourceReader.pollNext(IncrementalSourceReader.java:119)
at org.apache.seatunnel.engine.server.task.flow.SourceFlowLifeCycle.collect(SourceFlowLifeCycle.java:159)
at org.apache.seatunnel.engine.server.task.SourceSeaTunnelTask.collect(SourceSeaTunnelTask.java:127)
at org.apache.seatunnel.engine.server.task.SeaTunnelTask.stateProcess(SeaTunnelTask.java:165)
at org.apache.seatunnel.engine.server.task.SourceSeaTunnelTask.call(SourceSeaTunnelTask.java:132)
at org.apache.seatunnel.engine.server.TaskExecutionService$BlockingWorker.run(TaskExecutionService.java:694)
at org.apache.seatunnel.engine.server.TaskExecutionService$NamedTaskWrapper.run(TaskExecutionService.java:1023)
at org.apache.seatunnel.api.tracing.MDCRunnable.run(MDCRunnable.java:43)
at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:750)
Caused by: java.lang.RuntimeException: SplitFetcher thread 0 received unexpected exception while polling the records
at org.apache.seatunnel.connectors.seatunnel.common.source.reader.fetcher.SplitFetcher.runOnce(SplitFetcher.java:165)
at org.apache.seatunnel.connectors.seatunnel.common.source.reader.fetcher.SplitFetcher.run(SplitFetcher.java:81)
... 5 more
Caused by: org.apache.kafka.connect.errors.ConnectException: An exception occurred in the change event producer. This connector will be stopped.
at io.debezium.pipeline.ErrorHandler.setProducerThrowable(ErrorHandler.java:50)
at io.debezium.connector.oracle.logminer.LogMinerStreamingChangeEventSource.execute(LogMinerStreamingChangeEventSource.java:267)
at org.apache.seatunnel.connectors.seatunnel.cdc.oracle.source.reader.fetch.logminer.OracleRedoLogFetchTask$RedoLogSplitReadTask.execute(OracleRedoLogFetchTask.java:147)
at org.apache.seatunnel.connectors.seatunnel.cdc.oracle.source.reader.fetch.logminer.OracleRedoLogFetchTask.execute(OracleRedoLogFetchTask.java:73)
at org.apache.seatunnel.connectors.cdc.base.source.reader.external.IncrementalSourceStreamFetcher.lambda$submitTask$0(IncrementalSourceStreamFetcher.java:107)
... 5 more
Caused by: io.debezium.DebeziumException: Online REDO LOG files or archive log files do not contain the offset scn 2509495408. Please perform a new snapshot.
at io.debezium.connector.oracle.logminer.LogMinerStreamingChangeEventSource.execute(LogMinerStreamingChangeEventSource.java:171)
... 8 more
4.经过查询镜像库SCN号已经是29开头了,但是seatunel抽取的数据2509495408,导致seatunel同步进程无法正常启动
<img width="860" height="499" alt="Image" src="https://github.com/user-attachments/assets/9fac20af-da11-4db7-a391-d38230449578" />
--
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: commits-unsubscribe@seatunnel.apache.org.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "beta-zhou (via GitHub)" <gi...@apache.org>.
beta-zhou closed issue #11965: During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail
URL: https://github.com/apache/seatunnel/issues/11965
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] seatunel抽取Oracle数据库过程中,SCN后往回抽取,导致seatunel同步进程失败 [seatunnel]
Posted by "beta-zhou (via GitHub)" <gi...@apache.org>.
beta-zhou commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5420735930
这个问题已经反复多次,大概率是个BUG。特此提出问题
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "beta-zhou (via GitHub)" <gi...@apache.org>.
beta-zhou closed issue #11965: During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail
URL: https://github.com/apache/seatunnel/issues/11965
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "beta-zhou (via GitHub)" <gi...@apache.org>.
beta-zhou commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5452923938
Hi,
The tasks created on August 27 can run normally, but the new tasks need to reinitialize Doris data, and the Seatunnel extraction process needs to resync all the mirror database data to recover.
However, I think that during the rebuilding of the OGG process, the mirror database itself is running normally and the SCN numbers are progressing properly. In theory, Seatunnel should be able to read the SCN numbers normally. For the Seatunnel --> Doris extraction process, initializing the mirror database data doesn’t involve manually stopping the Seatunnel process. Also, the operation of initializing the mirror database (Oracle impdp) is essentially just executing DML operations.
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "beta-zhou (via GitHub)" <gi...@apache.org>.
beta-zhou commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5469451838
I think the most effective way might be for you to set up a test environment according to the scenario I described to reproduce the issue, that might be more effective.
However, I will follow your suggestion and do a detailed retest and record it. I'll close this issue for now. Thanks.
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "beta-zhou (via GitHub)" <gi...@apache.org>.
beta-zhou commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5465790802
Hi,
Maybe I didn't explain the configuration information I sent earlier—that's my fault. These configurations are for our production environment, and for security reasons, I made some changes to the important data, like IP addresses and database table info.
Based on your reply, can I understand it like this: when Seatunnel starts syncing for the first time, it records an SCN number where it starts extracting, and after running for a while, if the source is reset, Seatunnel will start extracting again from the SCN number recorded at the first start?
If that's the case, is there any command or debug mode to find out this SCN number where extraction began? We can use that to run the verification again, because this issue is easy to reproduce, and we've seen it happen multiple times already, which has caused quite a bit of impact on our downstream systems.
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "SEZ9 (via GitHub)" <gi...@apache.org>.
SEZ9 commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5469327500
Thanks for clarifying that the config values were modified for security. Just to be safe: the earlier note about rotating credentials only applies if the `password` values are real — if you changed those too, no action is needed.
On your question, a small correction to the mental model: on restart, SeaTunnel does not go back to the SCN recorded at the very first start. A fresh submission derives its startup position from `startup.mode` (defaulting to `initial` in your config), but a restarted job restores the checkpointed startup offset and continues from that persisted SCN rather than re-evaluating `startup.mode` against the current database state. So the resume point is the last checkpointed position — either way it is a stored SCN, not one recomputed after the source reset.
That matches what you observed: the old task kept failing after the OGG/mirror rebuild, while the task recreated on August 27 runs normally, because there is currently no automatic "discard old checkpoint and restart from fresh Oracle state" recovery path.
I don't have a verified command or debug flag to point you to for extracting the restored SCN, so I'd rather not guess. Since you can reproduce this easily, the evidence that would move this from expected restore semantics to a confirmed bug is still:
1. the exact restored SCN the failing old task resumes from;
2. verification that this SCN is still valid and available in the mirror after the OGG rebuild;
3. a failing log from a fresh job with no restored state (so far the fresh job is healthy);
4. confirmation of whether any old checkpoint/savepoint directory was reused when jobs were recreated.
If you can capture the full logs from the failing restarted job on your next reproduction, that would help us locate the restored SCN and confirm whether it was genuinely still valid when the restore failed.
<!-- streview-comment:676 -->
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "DanielLeens (via GitHub)" <gi...@apache.org>.
DanielLeens commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5453915372
Thanks, this new data is the most important part of the thread.
I rechecked the current Oracle-CDC path on `dev` (`1ff72ceb3d2a823a0fca64ec009984d75e846c28`). A fresh source derives its startup configuration from `startup.mode`, but a restored enumerator reuses the checkpointed startup offset instead of recomputing from the current database state. The base CDC assigner explicitly carries that restored startup offset forward across recovery.
That matches what you observed:
1. the new task created on August 27, 2026 runs normally;
2. the restarted old task keeps following its previously persisted SCN path;
3. SeaTunnel does not currently have an automatic "discard old checkpoint and restart from fresh Oracle state" recovery path for this case.
So, from the evidence in this issue, I still do not see proof that SeaTunnel moved the SCN backward by itself. The stronger explanation is that the persisted resume point became unusable for the rebuilt upstream capture path, while a fresh job on the rebuilt mirror is healthy.
If you still think there is a product bug beyond that recovery limitation, the missing evidence would be:
1. the exact restored SCN used by the failing old task;
2. proof that the same SCN is still valid and available after the OGG / mirror rebuild;
3. a failing log from a fresh job with no restored state.
Without that, this looks closer to expected restore semantics after an upstream reset than to a SeaTunnel CDC bug.
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "DanielLeens (via GitHub)" <gi...@apache.org>.
DanielLeens commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5438932197
Thanks for adding the concrete config and restart history.
I rechecked the current Oracle-CDC restore path on `apache/dev` (`e6caf5a4b337`). With this config, `startup.mode` is not set, so a fresh submission defaults to `initial`. But when the old job is restarted from its saved state, the CDC enumerator restores the checkpointed startup offset and continues from that stored SCN rather than re-evaluating `startup.mode`. That matches the error you saw: the old job keeps trying to resume from a stale SCN after the upstream OGG / mirror rebuild.
So, from the evidence in this thread, I still do not see proof that SeaTunnel moved the SCN backward by itself. The stronger explanation is:
1. the original streaming job kept an old checkpointed SCN;
2. after the upstream mirror / OGG path was rebuilt, that saved resume point was no longer a valid place to continue from;
3. recreating the job as a new sync task was the correct recovery step, because that clears the old restored offset path.
The most important remaining check is: does the newly created job run normally on the rebuilt mirror, or does it still repeat / fail from a fresh job with no restored state? If the fresh job is healthy, this issue is much closer to an expected recovery-after-upstream-reset case than to a SeaTunnel product bug.
If you continue investigating, please add:
1. whether the newly created job after August 27, 2026 runs normally or still duplicates / fails;
2. if it still fails, the new failing log from that fresh job, not the old restored one;
3. whether any old checkpoint/savepoint directory was reused when the job was recreated.
One more note: if the `password` values in the pasted config are real rather than placeholders, please rotate and redact them from the issue.
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "DanielLeens (via GitHub)" <gi...@apache.org>.
DanielLeens commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5425086224
Thanks for the detailed background and log.
From the current Oracle-CDC behavior, this error usually means SeaTunnel / Debezium is trying to resume from an old SCN that is no longer present in the Oracle redo or archive logs after the upstream OGG / mirror pipeline was rebuilt. In other words, `Online REDO LOG files or archive log files do not contain the offset scn ... Please perform a new snapshot.` points to an invalid resume position; by itself it does not prove that SeaTunnel moved the SCN backward.
On current SeaTunnel, Oracle-CDC supports `startup.mode = initial | latest | timestamp | specific`, and `startup.specific-offset.scn` must still exist in the selected log-mining backend. Also, when a CDC job is restored from checkpoint or savepoint, it resumes from the checkpointed offset path rather than starting a fresh snapshot automatically.
Before maintainers can call this a product bug, could you please add:
1. the exact SeaTunnel version and full Oracle-CDC source config, especially `startup.mode`, any `startup.specific-offset.scn`, heartbeat settings, and whether checkpointing / exactly-once is enabled;
2. whether this job was restarted from checkpoint / savepoint or submitted fresh after the OGG rebuild;
3. whether the OGG rebuild reinitialized the mirror from an earlier point or recreated redo / archive history so the old SCN disappeared;
4. whether restarting with `startup.mode=initial`, or from a new valid SCN / timestamp on the rebuilt mirror, resolves the failure.
If the old SCN is gone after the upstream rebuild, the immediate recovery path is usually a fresh snapshot (`startup.mode=initial`) or a controlled restart from a valid current SCN / timestamp on the rebuilt mirror.
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "beta-zhou (via GitHub)" <gi...@apache.org>.
beta-zhou commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5426480650
> Thanks for the detailed background and log.
>
> From the current Oracle-CDC behavior, this error usually means SeaTunnel / Debezium is trying to resume from an old SCN that is no longer present in the Oracle redo or archive logs after the upstream OGG / mirror pipeline was rebuilt. In other words, `Online REDO LOG files or archive log files do not contain the offset scn ... Please perform a new snapshot.` points to an invalid resume position; by itself it does not prove that SeaTunnel moved the SCN backward.
>
> On current SeaTunnel, Oracle-CDC supports `startup.mode = initial | latest | timestamp | specific`, and `startup.specific-offset.scn` must still exist in the selected log-mining backend. Also, when a CDC job is restored from checkpoint or savepoint, it resumes from the checkpointed offset path rather than starting a fresh snapshot automatically.
>
> Before maintainers can call this a product bug, could you please add:
>
> 1. the exact SeaTunnel version and full Oracle-CDC source config, especially `startup.mode`, any `startup.specific-offset.scn`, heartbeat settings, and whether checkpointing / exactly-once is enabled;
> 2. whether this job was restarted from checkpoint / savepoint or submitted fresh after the OGG rebuild;
> 3. whether the OGG rebuild reinitialized the mirror from an earlier point or recreated redo / archive history so the old SCN disappeared;
> 4. whether restarting with `startup.mode=initial`, or from a new valid SCN / timestamp on the rebuilt mirror, resolves the failure.
>
> If the old SCN is gone after the upstream rebuild, the immediate recovery path is usually a fresh snapshot (`startup.mode=initial`) or a controlled restart from a valid current SCN / timestamp on the rebuilt mirror.
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "beta-zhou (via GitHub)" <gi...@apache.org>.
beta-zhou commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5426654826
Hello, thanks for your reply.
The additional information needed is:
1. The OGG mirror database synchronization process is rebuilt by exporting data from production via Oracle's expdp and importing into the mirror database using impdp. During the impdp import into the mirror database, it essentially executes insert operations and does not reset the database's SCN; the SCN keeps increasing.
2. The mirror database has not cleaned or rebuilt the redo log, and the archived logs at the time when the Seatunnel process reported errors have also not been cleaned.
3. The wrong task is due to the original process failing, and even if the maintenance staff tried to restart it, it couldn't recover.
4. In the end, the maintenance staff deleted the original seatunel process and restored it by creating a new seatunel sync process.
5. The specific configuration details will be provided later.
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "beta-zhou (via GitHub)" <gi...@apache.org>.
beta-zhou commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5436256346
版本:apache-seatunnel-2.3.12
env{
execution.parallelism = 1
job.mode = "STREAMING"
job.name = "bike.bike_info"
checkpoint.interval =1000
}
source {
Oracle-CDC {
plugin_output = "tables14"
username = "logminer_user"
password = "pwd$123"
database-names = ["SJZTNCDB"]
schema-names = ["TESTDBO"]
table-names = ["SJZTNCDB.TESTDBO.bike_info"]
connect.timeout.ms = 120000
connect.max-retries = 30
base-url = "jdbc:oracle:thin:@192.168.56.10:1522:SJZTNCDB"
}
}
transform {
Sql {
plugin_input = "tables14"
query = "select *,NOW() as SEATUNNEL_UPDATE_TIME from tables14"
plugin_output = "tables14_tf"
}
}
sink {
Doris {
plugin_input = "tables14_tf"
fenodes = "192.168.56.20:8030"
username = root
password = "pwd$123"
database = "bike"
sink.enable-2pc = "false"
sink.label-prefix = "bike_14_TABLE1"
sink.enable-delete = "true"
doris.config = {
format="json"
read_json_by_line="true"
}
#save_mode_create_template = """CREATE TABLE IF NOT EXISTS `${database}`.`${table_name}` (${rowtype_fields}) ENGINE = OLAP UNIQUE KEY (${rowtype_primary_key}) DISTRIBUTED BY HASH (${rowtype_primary_key}) BUCKETS 10 PROPERTIES ("replication_num" = "3");"""
}
}
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "beta-zhou (via GitHub)" <gi...@apache.org>.
beta-zhou commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5469434102
Hi,
Maybe my problem description wasn't clear. What I actually meant is that when I rebuilt the OGG sync process, Seatunnel picked up the old SCN number from before, which caused the Seatunnel extraction process to fail. It's certain that Seatunnel really can't change the SCN number of the database. ...
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] During the seatunel extraction from the Oracle database, SCN was extracted backward, causing the seatunel sync process to fail [seatunnel]
Posted by "DanielLeens (via GitHub)" <gi...@apache.org>.
DanielLeens commented on issue #11965:
URL: https://github.com/apache/seatunnel/issues/11965#issuecomment-5468621798
Thanks, this follow-up clarifies the operational question.
I rechecked the current Oracle-CDC restore path on `dev` again. The important distinction is:
1. for a **fresh** job, `startup.mode` is resolved when the source is created;
2. for a **restored** job, SeaTunnel reuses the offset that was already checkpointed for that job state, rather than asking Oracle for a brand new start SCN on every restart.
So the behavior is not "keep the SCN from the very first startup forever". It is closer to "keep resuming from the last persisted checkpoint/savepoint offset for that existing job state".
On the current code path, I do **not** see a dedicated CLI or REST command that prints that restored Oracle SCN directly from saved state. The most practical verification path available today is:
1. reproduce on the **old restored job**, not the newly created fresh job;
2. enable DEBUG logging for the Oracle CDC / LogMiner path and capture the lines that print `Offset SCN` and `Offset Commit SCN` during the failing restart;
3. compare that restored SCN with the earliest SCN still available in the redo/archive window after the OGG rebuild.
That evidence would answer the key question here: whether the restarted old job is simply resuming from an old persisted offset that is no longer valid after the upstream reset, or whether a fresh job with no reused state can also fail the same way.
So if you continue this investigation, the most useful next update would be:
1. the DEBUG log from the failing **restored old job** showing the `Offset SCN`;
2. confirmation of whether that SCN still exists in the Oracle redo/archive range after the OGG rebuild;
3. confirmation that the fresh job created on August 27, 2026 stays healthy over time without reusing the old checkpoint/savepoint state.
With those three points, maintainers can verify whether this is mainly an offset-observability gap around restore semantics, or a deeper Oracle-CDC bug.
--
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: commits-unsubscribe@seatunnel.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org