Fwd: BUG #17467: Perf degradation after switching to latest jdbc drivers - Mailing list pgsql-jdbc

From Benjamin Leis
Subject Fwd: BUG #17467: Perf degradation after switching to latest jdbc drivers
Date
Msg-id CADP9qHvLD53E-Ff4dh8f0WGCz+HS_CHHgt08+AHXOZQCFXHj4g@mail.gmail.com
Whole thread Raw
Responses Re: BUG #17467: Perf degradation after switching to latest jdbc drivers  (Dave Cramer <davecramer@postgres.rocks>)
List pgsql-jdbc

Forwarding this to the jdbc list.

Bug reference:      17467
Logged by:          Benjamin Leis
Email address:      benleis1@gmail.com
PostgreSQL version: 13.4
Operating system:   Linux
Description:       

Due to the CVE-2022-21724  we upgraded our driver from 42.2.22 to 43.3.3.
We've now started to see a large perf degradation during initial opens to the
database.

With a small perf test that just opens and tries a "select 1"  we're getting
a ~2 - 2.5x slowdown on the connect times.  Is this a known issue or
something that we can work around?

43.2.25 driver
INFO - New connection test
INFO - Open time:  25.00 Call time: 14.00
INFO - Open time:  24.00 Call time: 1.00
INFO - Open time:  19.00 Call time: 1.00
INFO - Open time:  16.00 Call time: 1.00
INFO - Open time:  26.00 Call time: 4.00
INFO - Open time:  20.00 Call time: 6.00
INFO - Open time:  20.00 Call time: 1.00
INFO - Open time:  16.00 Call time: 0.00
INFO - Open time:  19.00 Call time: 1.00
INFO - Open time:  19.00 Call time: 1.00

43.3.3 driver
INFO - New connection test
INFO - Open time: 57.00 Call time: 14.00
INFO - Open time: 58.00 Call time: 1.00
INFO - Open time: 62.00 Call time: 1.00
INFO - Open time: 54.00 Call time: 1.00
INFO - Open time: 59.00 Call time: 1.00
INFO - Open time: 56.00 Call time: 0.00
INFO - Open time: 59.00 Call time: 1.00
INFO - Open time: 61.00 Call time: 1.00
INFO - Open time: 57.00 Call time: 1.00
INFO - Open time: 58.00 Call time: 1.00

pgsql-jdbc by date:

Previous
From: "David G. Johnston"
Date:
Subject: Re: Why does the prepareThreshold=0 (to cover pgbouncer transaction mode) disables protocol binary transfers ? (w/ PoC patch and measurements)
Next
From: Andrew Dunstan
Date:
Subject: Re: Why does the prepareThreshold=0 (to cover pgbouncer transaction mode) disables protocol binary transfers ? (w/ PoC patch and measurements)