1. Should we assign uid and name to the sources and sinks too ? >> If the sources/sinks have used state, you should assign uid for them. This is usually true for sources.
2. What are the pros and cons of adding uid to sources and sinks ? >> I'm not seeing the cons for assigning uid to sources and sinks. So I guess assigning the uids for sources/sinks is always a good practice. 3. The sinks have uid and hashUid - which is the preferred attribute to use for allowing job restarts ? >> Could you see if this could answer you question: https://stackoverflow.com/questions/46112142/apache-flink-set-operator-uid-vs-uidhash 4. If sink and sources uid are not provided in the application, can they still maintain state across job restarts from checkpoints ? >> It depends on whether the sources/sinks uses state. I think most sources use state to maintaining the read offset. 5. Can the sinks and sources without uid restart from savepoints ? >> The same as above. 6. The data streams have an attribute id - How is this generated and can this be used for creating a uid for the sink ? >> Not sure what do you mean by "attribute id". Could you give some more detailed information about it? Regards, Dian On Fri, Nov 22, 2019 at 6:27 PM M Singh <mans2si...@yahoo.com> wrote: > > Hi Folks - Please let me know if you have any advice on the best practices > for setting uid for sources and sinks. Thanks. Mans > On Thursday, November 21, 2019, 10:10:49 PM EST, M Singh < > mans2si...@yahoo.com> wrote: > > > Hi Folks: > > I am assigning uid and name for all stateful processors in our application > and wanted to find out the following: > > 1. Should we assign uid and name to the sources and sinks too ? > 2. What are the pros and cons of adding uid to sources and sinks ? > 3. The sinks have uid and hashUid - which is the preferred attribute to > use for allowing job restarts ? > 4. If sink and sources uid are not provided in the application, can they > still maintain state across job restarts from checkpoints ? > 5. Can the sinks and sources without uid restart from savepoints ? > 6. The data streams have an attribute id - How is this generated and can > this be used for creating a uid for the sink ? > > Thanks for your help. > > Mans >
