Hi Andrey, yes, I think flip 34 helps with this use case.
https://flink.apache.org/news/2019/08/22/release-1.9.0.html#stop-with-savepoint-flip-34 Thanks, Chen On Tue, Nov 26, 2019 at 7:31 AM Andrey Zagrebin <azagre...@apache.org> wrote: > Hi Chen, > > if I understand correctly, your question is not about the state with TTL, > correct me if I am wrong. > > We could consider adding this feature to Flink: run x time => save point > => shutdown job but it seems to me that it is something quite application > specific, not sure how high priority it will get in community. > > Would it work for you to schedule an automated cron script to check your > running job and do those actions over CLI or REST API? > > Cheers, > Andrey > > On Mon, Nov 25, 2019 at 7:27 PM Yun Tang <myas...@live.com> wrote: > > > Hi Chen > > > > There existed several problems currently: > > 1. State TTL cannot be applied to Flink SQL > > 2. No job-level state TTL > > 3. Window state should not be limited to TTL if existed a job-level state > > > > If your goal is to achieve #2, we could introduce a configuration via > > flink-conf.yaml so that we could also achieve #1 but leave window state > not > > limited to TTL. > > I think this deserves an issue to track this request. > > > > Best > > Yun Tang > > > > On 11/25/19, 8:04 AM, "Chen Qin" <qinnc...@gmail.com> wrote: > > > > I am a bit confused. Job level TTL is different from state TTL > despite > > share same name. > > > > What we want to achieve in Job TTL is basically "run x time => save > > point > > => shutdown job" > > > > > > On Sun, Nov 24, 2019 at 9:57 AM Yun Tang <myas...@live.com> wrote: > > > > > Hi Chen > > > > > > Currently, state TTL setting is per-state scoped, and there is no > > > job-level TTL setting so far. Do you want similar settings in > > flink-conf to > > > enable Flink SQL could also benefit from state TTL? > > > > > > Best > > > Yun Tang > > > > > > On 11/25/19, 12:16 AM, "Chen Qin" <qinnc...@gmail.com> wrote: > > > > > > Hi there, > > > > > > We have use case of allowing ad hoc user submit a Flink job > > (mostly > > > SQL) > > > and user come back and verify results after period of time (eg > > 24hours) > > > without manual intervention. The idea is to have something like > > set job > > > level TTL configuration to 24hours. Jobmanager will honor > > 24hours of > > > instrument save point and proper shutdown of flink job. > > > > > > Just want to know if we have job level TTL setting or parity > > features > > > JIRA. > > > > > > > > > Best, > > > Chen > > > > > > > > > > > > > > > >
