When you modify a job by removing a stateful operator, then during a restart when Flink tries to restore the state, it will complain that the snapshot contains state that can not be restored.
The solution to this is to restart from the savepoint (or retained checkpoint), specifying that you want to allow non-restored state [1]: ./bin/flink run -s <savepointPath> --allowNonRestoredState BTW, you may run into further problems if you haven't assigned UIDs to your stateful operators. [2] [1] https://ci.apache.org/projects/flink/flink-docs-stable/ops/cli.html#restore-a-savepoint [2] https://ci.apache.org/projects/flink/flink-docs-stable/ops/state/savepoints.html#assigning-operator-ids On Tue, Jul 28, 2020 at 4:11 PM ApoorvK <apoorv.upadh...@razorpay.com> wrote: > Hi All, > > I have multiple steam in flink job which contains different state such as > ValueState or MapState. > But I now I need to remove one stream having specific (UID,NAME) from the > JOB. > > If I remove it I face issue while restoration stating operator does not > exists. > > I was using BucketSink for sinking data to HDFS with let say uid->hdfs-sink > name->hdf-sink. > Now I have changed it to sink into S3 using StreamingFile sink keeping > uid->hdfs-sink name->hdf-sink same as restoration will fail , but still it > is failing when I restore from savepoint , this is probably due to > BucketSink to StreamingFile change. > > How can I achieve this then I need to remove HDFS and keep s3 > > > > -- > Sent from: > http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/ >
