rust/neon
1
0
Fork 0
mirror of https://github.com/neondatabase/neon.git synced 2026-06-03 13:30:38 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update pgxn/neon/pagestore_smgr.c

Browse Source 
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
This commit is contained in:
Konstantin Knizhnik
2022-12-13 12:40:25 +02:00
committed by GitHub
parent f6f87bb6f3
commit 407d6b120b
1 changed files with 20 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
pgxn/neon/pagestore_smgr.c
View File

@@ -681,7 +681,26 @@ prefetch_do_request(PrefetchRequest *slot, bool *force_latest, XLogRecPtr *force
Assert(slot->my_ring_index == MyPState->ring_unused);
page_server->send((NeonRequest *) &request);
/* Remember request LSN for the prefetched page to avoid false pefetch invalidations */
/*
* Remember request LSN in the last-written LSN cache to avoid false
* prefetch invalidations.
*
* Imagine what would happen without this, when you perform a large
* sequential scan with UPDATE. The sequential scan issues a prefetch
* request for each page in order, and every page is also dirtied. On
* each page, the oldest page in the last-written LSN cache is evicted,
* which advances the global last-written LSN. The pages being scanned are
* not in the last-written cache, so each prefetch request will use the
* global last-written LSN in the request and memorize that in the
* slot. However, when we receive the response to the prefetch request,
* the global last-written LSN has already moved forwards, and the
* cross-check we make that the last-written LSN matches will fail, and we
* discard the prefetched response unnecessary.
*
* Inserting the LSN we use in the prefetch request to the last-written LSN
* cache avoids that problem. With that, we will use the cached value in
* the cross-check, instead of the more recent global last-written LSN value.
*/
SetLastWrittenLSNForBlock(
request.req.lsn,
slot->buftag.rnode,