Hi hackers. I propose a small optimization can be added to the tablesync replication code.
This proposal (and simple patch) was first discussed here [1]. Basic idea is the tablesync could/should detect if there is anything to do *before* it enters the apply main loop. Calling process_sync_tables() before the apply main loop offers a quick way out, so the message handling will not be unnecessarily between workers. This will be a small optimization. But also, IMO this is a more natural separation of work. E.g tablesync worker will finish when the table is synced - not go one extra step... ~~ This patch was already successfully used for several versions (v43-v50) of another 2PC patch [2], but it was eventually removed from there because, although it has its own independent value, it was not required for that patch series [3]. ---- [1] https://www.postgresql.org/message-id/CAHut%2BPtjk-Qgd3R1a1_tr62CmiswcYphuv0pLmVA-%2B2s8r0Bkw%40mail.gmail.com [2] https://www.postgresql.org/message-id/flat/CAHut%2BPsd5nyg-HG6rGO2_5jzXuSA1Eq5%2BB5J2VJo0Q2QWi-1HQ%40mail.gmail.com#1c268eeee3756b32e267d96b7177ba95 [3] https://www.postgresql.org/message-id/CAA4eK1Jxu-3qxtkfA_dKoquQgGZVcB%2Bk9_-yT5%3D9GDEW84TF%2BA%40mail.gmail.com Kind Regards, Peter Smith. Fujitsu Australia
