db-koji01: more tuning

Tweak some of the knobs to try and get things more happy with the heavy
load it's under. Note that cpus change will need a reboot before it's
active. The other changes have all been applied and it seems to be doing
a little better with this. Perhaps with some more cpus we can get it
stable (but yet not swamping i/o).

Signed-off-by: Kevin Fenzi <kevin@scrye.com>
This commit is contained in:
Kevin Fenzi 2023-11-09 10:09:24 -08:00
parent 21b43ab931
commit 9314b70862

View file

@ -12,7 +12,7 @@ db_backup_dir: ['/backups']
dbs_to_backup:
- koji
effective_cache_size: "100GB"
effective_io_concurrency: 0
effective_io_concurrency: 1
eth0_ipv4_gw: 10.3.169.254
eth0_ipv4_ip: 10.3.169.103
# kernel SHMMAX value
@ -36,8 +36,9 @@ nrpe_procs_crit: 700
# Large updates pushes cause lots of db threads doing the tag moves, so up this from default.
#
nrpe_procs_warn: 600
num_cpus: 48
shared_buffers: "30GB"
num_cpus: 64
shared_buffers: "48GB"
swap: false
temp_buffers: "32MB"
vmhost: bvmhost-x86-01.iad2.fedoraproject.org
volgroup: /dev/vg_guests