On Mon, Dec 5, 2022 at 3:41 PM Dilip Kumar <dilipbal...@gmail.com> wrote: > > On Mon, Dec 5, 2022 at 9:21 AM Dilip Kumar <dilipbal...@gmail.com> wrote: > > > > On Mon, Dec 5, 2022 at 8:59 AM Amit Kapila <amit.kapil...@gmail.com> wrote: > > > > > > On Sun, Dec 4, 2022 at 5:14 PM houzj.f...@fujitsu.com > > > <houzj.f...@fujitsu.com> wrote: > > > > > > > > On Saturday, December 3, 2022 7:37 PM Amit Kapila > > > > <amit.kapil...@gmail.com> wrote: > > > > > Apart from the above, I have slightly adjusted the comments in the > > > > > attached. Do > > > > > let me know what you think of the attached. > > > > > > > > Thanks for updating the patch. It looks good to me. > > > > > > > > > > I feel the function name ReorderBufferLargestTopTXN() is slightly > > > misleading because it also checks some of the streaming properties > > > (like whether the TXN has partial changes and whether it contains any > > > streamable change). Shall we rename it to > > > ReorderBufferLargestStreamableTopTXN() or something like that? > > > > Yes that makes sense > > I have done this change in the attached patch. > > > > The other point to consider is whether we need to have a test case for > > > this patch. I think before this patch if the size of DDL changes in a > > > transaction exceeds logical_decoding_work_mem, the empty streams will > > > be output in the plugin but after this patch, there won't be any such > > > stream. > > I tried this test, but I think generating 64k data with just CID > messages will make the test case really big. I tried using multiple > sessions such that one session makes the reorder buffer full but > contains partial changes so that we try to stream another transaction > but that is not possible in an automated test to consistently generate > the partial change. >
I also don't see a way to achieve it in an automated way because both toast and speculative inserts are part of one statement, so we need a real concurrent test to make it happen. Can anyone else think of a way to achieve it? > I think we need something like this[1] so that we can better control > the streaming. > +1. The additional advantage would be that we can generate parallel apply and new streaming tests with much lesser data. Shi-San, can you please start a new thread for the GUC patch proposed by you as indicated by Dilip? -- With Regards, Amit Kapila.
