On Wed, Jun 19, 2024 at 05:14:50AM +0000, Hayato Kuroda (Fujitsu) wrote: > I have an unclear point. According to the comment atop GetInsertRecPtr(), it > just > returns the approximated value - the position of the last full WAL page [1]. > If there is a continuation WAL record which across a page, will it return the > halfway point of the WAL record (end of the first WAL page)? If so, the > proposed > fix seems not sufficient. We have to point out the exact the end of the > record.
Yeah, that a thing of the patch I am confused with. How are we sure that this is the correct LSN to rely on? If that it the case, the patch does not offer an explanation about why it is better. WalSndWaitForWal() is called only in the context of page callback for a logical WAL sender. Shouldn't we make the flush conditional on what's stored in XLogReaderState.missingContrecPtr? Aka, if we know that we're in the middle of the decoding of a continuation record, we should wait until we've dealt with it, no? In short, I would imagine that WalSndWaitForWal() should know more about XLogReaderState is doing. But perhaps I'm missing something. -- Michael
signature.asc
Description: PGP signature