From: | Daniel Gustafsson <daniel(at)yesql(dot)se> |
---|---|
To: | Peter Eisentraut <peter(at)eisentraut(dot)org> |
Cc: | Nathan Bossart <nathandbossart(at)gmail(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
Subject: | Re: Reducing connection overhead in pg_upgrade compat check phase |
Date: | 2023-10-27 13:20:09 |
Message-ID: | 7B104870-5A35-4FC1-AF4F-7E9C8EDBDF18@yesql.se |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Attached is a v10 rebase of this patch which had undergone significant bitrot
due to recent changes in the pg_upgrade check phase. This brings in the
changes into the proposed structure without changes to queries, with no
additional changes to the proposed functionality.
Testing with a completely empty v11 cluster fresh from initdb as the old
cluster shows a significant speedup (averaged over multiple runs, adjusted for
outliers):
patched: 53.59ms (52.78ms, 52.49ms, 55.49ms)
master : 125.87ms (125.23 ms, 125.67ms, 126.67ms)
Using a similarly empty cluster from master as the old cluster shows a smaller
speedup, which is expected since many checks only run for older versions:
patched: 33.36ms (32.82ms, 33.78ms, 33.47ms)
master : 44.87ms (44.73ms, 44.90ms 44.99ms)
The latter case is still pretty interesting IMO since it can speed up testing
where every millisecond gained matters.
--
Daniel Gustafsson
Attachment | Content-Type | Size |
---|---|---|
v10-0001-pg_upgrade-run-all-data-type-checks-per-connecti.patch | application/octet-stream | 41.4 KB |
From | Date | Subject | |
---|---|---|---|
Next Message | David Steele | 2023-10-27 13:31:10 | Re: Requiring recovery.signal or standby.signal when recovering with a backup_label |
Previous Message | Dean Rasheed | 2023-10-27 13:05:21 | Re: btree_gin: Incorrect leftmost interval value |