Jason,
This is excellent! I, for one, am excited that someone is looking into
this subject deeply. I've been wanting to migrate to a newer version of
PostgreSQL from quite some time. We have some test machines running
Evergreen on PG 10 but they are not getting the thorough testing that
I'd like to be certain it's ready for production. So, thank you. Thank
you very much!
In order to configure PG to use 40% of the server's hardware, are you
focusing on settings?:
|work_mem max_connections effective_cache_size default_statistics_target
constraint_exclusion checkpoint_completion_target wal_buffers
shared_buffers maintenance_work_mem|
And kernel setting:
|vm.overcommit_memory |
Or something else?
I think it's a good idea to test the same version of Evergreen, like you
said. And the latest version makes the most sense.
-Blake-
Conducting Magic
Can consume data in any format
MOBIUS
On 7/21/2021 8:53 AM, Jason Stephenson via Evergreen-dev wrote:
Hello, all.
As some of you are aware, I have been running Evergreen against
various PostgreSQL releases on an old email server that has 128 GB of
RAM and 8TB of disk space. I currently have PostgreSQL versions 9.6
through 13 installed.
During my testing, I have identified a few areas where Evergreen
database performance is slower on newer Pg versions and other areas
where it is faster.
I am presently setting up to use the plprofiler extension
(https://github.com/bigsql/plprofiler) to test these areas, beginning
with those where I know that performance degrades in PostgreSQL
version greater than 10.
To make the test fair, I am loading a dump of CW MARS' production
database from Sunday, July 11, 2021 into the various Pg instances.
Since we are currently on Evergreen 3.5.3, I intend to also run a
database upgrade script on each instance.
I have a question regarding the database upgrade: Should I upgrade to
Evergreen 3.7.1 or to an arbitrary commit on the master branch, i.e.
the one that is current as of the time that I generate the db upgrade?
I also intend to use optimization settings on the database instance
being tested. I presently have 2 of the instances configured to use
roughly 40% of the server's capabilities and the other instances are
left at the default settings. I may continue with the 40% optimization
or go with settings that will utilize more of the hardware for just
the instance being tested and either shut off or run the other
instances with default settings.
If anyone has any additional ideas of things that should be tested or
would like to help devise a script to more thoroughly test the
Evergreen database, let me know.
Cheers,
Jason
_______________________________________________
Evergreen-dev mailing list
[email protected]
http://list.evergreen-ils.org/cgi-bin/mailman/listinfo/evergreen-dev
_______________________________________________
Evergreen-dev mailing list
[email protected]
http://list.evergreen-ils.org/cgi-bin/mailman/listinfo/evergreen-dev
