Re: [DISCUSS] FLIP-272: Generalized delegation token support

2022-11-22 Thread Gabor Somogyi
Hi Danny, Please find my answers inline. G On Mon, Nov 21, 2022 at 9:27 PM Danny Cranmer wrote: > Hey, > > > All the rest which is public can be considered accidental leak (for > example > "DelegationTokenManager" where Java doesn't support internal interfaces). > > If the classes are interna

Re: [DISCUSS] FLIP-272: Generalized delegation token support

2022-11-21 Thread Danny Cranmer
Hey, > All the rest which is public can be considered accidental leak (for example "DelegationTokenManager" where Java doesn't support internal interfaces). If the classes are internal, then they should be annotated @Internal. This is a Flink thing [1]. My explicit recommendation is to ensure all

Re: [DISCUSS] FLIP-272: Generalized delegation token support

2022-11-21 Thread Gabor Somogyi
Hi Danny, Thanks for spending your time to make the design better. My comments added inline. Please share your opinion or just correct me if I've misunderstood your suggestions. G On Mon, Nov 21, 2022 at 5:10 PM Danny Cranmer wrote: > Thanks for the FLIP Gabor. > > Generally looks good, have

Re: [DISCUSS] FLIP-272: Generalized delegation token support

2022-11-21 Thread Danny Cranmer
Thanks for the FLIP Gabor. Generally looks good, have a couple of comments. If I understand correctly we plan on pulling the current implementation apart and adding an extra layer of abstraction, to allow arbitrary implementations. I notice that most of the delegation token classes [1] are missin

Re: [DISCUSS] FLIP-272: Generalized delegation token support

2022-11-09 Thread Gabor Somogyi
OK, removed the fallback part. Thanks for the help! G On Wed, Nov 9, 2022 at 5:03 PM Őrhidi Mátyás wrote: > looks better! no further concerns > > Cheers, > Matyas > > On Mon, Nov 7, 2022 at 9:21 AM Gabor Somogyi > wrote: > > > Oh gosh, copied wrong config keys so fixed my last mail with gree

Re: [DISCUSS] FLIP-272: Generalized delegation token support

2022-11-09 Thread Őrhidi Mátyás
looks better! no further concerns Cheers, Matyas On Mon, Nov 7, 2022 at 9:21 AM Gabor Somogyi wrote: > Oh gosh, copied wrong config keys so fixed my last mail with green. > > On Mon, Nov 7, 2022 at 6:07 PM Gabor Somogyi > wrote: > > > Hi Matyas, > > > > In the meantime I was thinking about the

Re: [DISCUSS] FLIP-272: Generalized delegation token support

2022-11-07 Thread Gabor Somogyi
Oh gosh, copied wrong config keys so fixed my last mail with green. On Mon, Nov 7, 2022 at 6:07 PM Gabor Somogyi wrote: > Hi Matyas, > > In the meantime I was thinking about the per provider re-obtain feature > and here are my thoughts related that: > * I think it's a good feature in general but

Re: [DISCUSS] FLIP-272: Generalized delegation token support

2022-11-07 Thread Gabor Somogyi
Hi Matyas, In the meantime I was thinking about the per provider re-obtain feature and here are my thoughts related that: * I think it's a good feature in general but as mentioned I would add it in a separate FLIP * In case of Hadoop providers it just wouldn't work (HBase doesn't have end timestam

Re: [DISCUSS] FLIP-272: Generalized delegation token support

2022-11-07 Thread Gabor Somogyi
Hi Matyas, Thanks for your comments, answered inline. G On Mon, Nov 7, 2022 at 2:58 PM Őrhidi Mátyás wrote: > Hi Gabor, > > Thanks for driving this effort! A few thoughts on the topic: > - Could you please add a few examples of the delegation token providers we > expected to be added in the n

Re: [DISCUSS] FLIP-272: Generalized delegation token support

2022-11-07 Thread Őrhidi Mátyás
Hi Gabor, Thanks for driving this effort! A few thoughts on the topic: - Could you please add a few examples of the delegation token providers we expected to be added in the near future? Ideally these providers could be configured independently from each other. However the configuration defaults

[DISCUSS] FLIP-272: Generalized delegation token support

2022-11-07 Thread Gabor Somogyi
Hi team, Delegation token framework is going to be finished soon (added in FLIP-211 ). Previously there were concerns that the current implementation is bound to Hadoop an