Skip to content

[Backport perf-v16] ci(perf): add latency steady state perf test using latte #10908

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
fruch merged 1 commit into from
Jun 23, 2025
Merged

[Backport perf-v16] ci(perf): add latency steady state perf test using latte #10908

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
233 changes: 233 additions & 0 deletions ...gurations/performance/latte-perf-regression-latency-steady-state-custom-d1-workload1.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,233 @@
test_duration: 300
n_monitor_nodes: 1
n_db_nodes: 8
n_loaders: 2
stress_duration: 60

gce_instance_type_db: 'n2-highmem-32'
gce_instance_type_loader: 'e2-highcpu-32'

store_perf_results: false
use_prepared_loaders: false
use_hdrhistogram: true

user_prefix: 'perf-lat-regr-custom-d1-w1'

# NOTE: number of local SSDs which can be attached to the 'n2-highmem-32' instance type
# must be divisible by 4 (platform requirement).
gce_n_local_ssd_disk_db: 4
# NOTE: each local SSD on GCE has 375Gb, so PD size must match 'ssd-num'*'ssd-size' formula.
gce_pd_ssd_disk_size_db: 1500
gce_setup_hybrid_raid: true

use_preinstalled_scylla: true

scylla_d_overrides_files: [
'scylla-qa-internal/custom_d1/workload1/scylla.d/cpuset.conf',
'scylla-qa-internal/custom_d1/workload1/scylla.d/io.conf',
'scylla-qa-internal/custom_d1/workload1/scylla.d/io_properties.yaml',
]

append_scylla_yaml:
reader_concurrency_semaphore_cpu_concurrency: 10

latency_decorator_error_thresholds:
write:
"Steady State":
P90 write:
fixed_limit: 10
P99 write:
fixed_limit: 10
Throughput write:
best_pct: 20
read:
"Steady State":
P90 read:
fixed_limit: 12
P99 read:
fixed_limit: 12
Throughput read:
best_pct: 20

round_robin: true
# NOTE: stress commands were copied from the custom-d1/workload1 scenario located at
# test-cases/longevity/longevity-gce-custom-d1-workload1-hybrid-raid.yaml
prepare_write_cmd:
# NOTE: --duration in these commands is number of rows that will be written.
# Time gets specified with 's', 'm' or 'h' letters.
- >-
latte run --tag latte-prepare-01 --duration 82286400 --request-timeout 30 --retry-interval '2s,10s'
--threads 30 --concurrency 180 --rate 17000 --sampling 5s -P offset=0
--function t1__insert -P row_count=82286400
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
- >-
latte run --tag latte-prepare-02 --duration 82286400 --request-timeout 30 --retry-interval '2s,10s'
--threads 30 --concurrency 180 --rate 17000 --sampling 5s -P offset=82286400
--function t1__insert -P row_count=82286400
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
- >-
latte run --tag latte-prepare-03 --duration 82286400 --request-timeout 30 --retry-interval '2s,10s'
--threads 30 --concurrency 180 --rate 17000 --sampling 5s -P offset=0
--function t2__insert -P row_count=82286400
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
- >-
latte run --tag latte-prepare-04 --duration 82286400 --request-timeout 30 --retry-interval '2s,10s'
--threads 30 --concurrency 180 --rate 17000 --sampling 5s -P offset=82286400
--function t2__insert -P row_count=82286400
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
- >-
latte run --tag latte-prepare-05 --duration 82286400 --request-timeout 30 --retry-interval '2s,10s'
--threads 30 --concurrency 180 --rate 17000 --sampling 5s -P offset=0
--function t3__insert -P row_count=82286400
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
- >-
latte run --tag latte-prepare-06 --duration 82286400 --request-timeout 30 --retry-interval '2s,10s'
--threads 30 --concurrency 180 --rate 17000 --sampling 5s -P offset=82286400
--function t3__insert -P row_count=82286400
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
- >-
latte run --tag latte-prepare-07 --duration 82286400 --request-timeout 30 --retry-interval '2s,10s'
--threads 30 --concurrency 180 --rate 17000 --sampling 5s -P offset=0
--function t4__insert -P row_count=82286400
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
- >-
latte run --tag latte-prepare-08 --duration 82286400 --request-timeout 30 --retry-interval '2s,10s'
--threads 30 --concurrency 180 --rate 17000 --sampling 5s -P offset=82286400
--function t4__insert -P row_count=82286400
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
- >-
latte run --tag latte-prepare-09 --duration 82286400 --request-timeout 30 --retry-interval '2s,10s'
--threads 30 --concurrency 180 --rate 17000 --sampling 5s -P offset=0
--function t5__insert -P row_count=82286400
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
- >-
latte run --tag latte-prepare-10 --duration 82286400 --request-timeout 30 --retry-interval '2s,10s'
--threads 30 --concurrency 180 --rate 17000 --sampling 5s -P offset=82286400
--function t5__insert -P row_count=82286400
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn

stress_cmd:
# NOTE: 'latte' tag will be used by the log collector code.
# 01) T4F10 / t4__get_many_lte -> -r 51.5k (~1/2 from 103k) - part1
- >-
latte run --tag latte-main-01 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 180 --rate 51500
--function t4__get_many_lte -P row_count=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 02) T4F10 / t4__get_many_lte -> -r 51.5k (~1/2 from 103k) - part2
- >-
latte run --tag latte-main-02 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 180 --rate 51500
--function t4__get_many_lte -P row_count=82286400 -P offset=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 03) T4F6 / t4__get -> -r 29.5k (~1/2 from 59k) - part1
- >-
latte run --tag latte-main-03 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 120 --rate 29500
--function t4__get -P row_count=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 04) T4F6 / t4__get -> -r 29.5k (~1/2 from 59k) - part2
- >-
latte run --tag latte-main-04 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 120 --rate 29500
--function t4__get -P row_count=82286400 -P offset=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 05) T4F1 / t4__insert -> -r 15.5k (~1/2 from 31k) - part1
- >-
latte run --tag latte-main-05 --duration 60m --request-timeout 30 --retry-interval '1s,5s'
--sampling 5s --threads 30 --concurrency 120 --rate 15500
--function t4__insert -P row_count=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 06) T4F1 / t4__insert -> -r 15.5k (~1/2 from 31k) - part2
- >-
latte run --tag latte-main-06 --duration 60m --request-timeout 30 --retry-interval '1s,5s'
--sampling 5s --threads 30 --concurrency 120 --rate 15500
--function t4__insert -P row_count=82286400 -P offset=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 07) T4F9 / t4__get_many_lt -> -r 11k (~1/2 from 22k) - part1
- >-
latte run --tag latte-main-07 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 120 --rate 11000
--function t4__get_many_lt -P row_count=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 08) T4F9 / t4__get_many_lt -> -r 11k (~1/2 from 22k) - part2
- >-
latte run --tag latte-main-08 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 120 --rate 11000
--function t4__get_many_lt -P row_count=82286400 -P offset=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 09) T4F7 / t4__get_many_gt -> -r 3k (~1/2 from 6k) - part1
- >-
latte run --tag latte-main-09 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 120 --rate 3000
--function t4__get_many_gt -P row_count=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 10) T4F7 / t4__get_many_gt -> -r 3k (~1/2 from 6k) - part2
- >-
latte run --tag latte-main-10 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 120 --rate 3000
--function t4__get_many_gt -P row_count=82286400 -P offset=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 11) T4F2 / t4__update -> -r 2k (~1/2 from 4k) - part1
- >-
latte run --tag latte-main-11 --duration 60m --request-timeout 30 --retry-interval '1s,5s'
--sampling 5s --threads 30 --concurrency 120 --rate 2000
--function t4__update -P row_count=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 12) T4F2 / t4__update -> -r 2k (~1/2 from 4k) - part2
- >-
latte run --tag latte-main-12 --duration 60m --request-timeout 30 --retry-interval '1s,5s'
--sampling 5s --threads 30 --concurrency 120 --rate 2000
--function t4__update -P row_count=82286400 -P offset=82286400 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 13) T3F4/t3__get -> -r 500
- >-
latte run --tag latte-main-13 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 120 --rate 500
--function t3__get -P row_count=164572800 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 14) T5F3/t5__get -> -r 500
- >-
latte run --tag latte-main-14 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 120 --rate 500
--function t5__get -P row_count=164572800 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 15) T3F1/t3__insert , T5F1/t5__insert -> -r 500 (= ~ 1k)
- >-
latte run --tag latte-main-15 --duration 60m --request-timeout 30 --retry-interval '1s,5s'
--sampling 5s --threads 30 --concurrency 120 --rate 1000
--function t3__insert,t5__insert -P row_count=164572800 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 16) T1F1-2 , T2F1-2 , T3F2 , T4F3-5 , T5F2 -> -r 90 (= ~810)
- >-
latte run --tag latte-main-16 --duration 60m --request-timeout 30 --retry-interval '1s,5s'
--sampling 5s --threads 30 --concurrency 120 --rate 810
--function t1__insert,t1__update,t2__insert,t2__update,t3__update,t4__update_udt1,t4__update_udt2,t4__update_bool,t5__update
-P row_count=164572800 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 17) T1F3-5 , T4F8 -> -r 90 (= ~360)
- >-
latte run --tag latte-main-17 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 120 --rate 360
--function t1__count,t1__get,t1__get_all,t4__get_many_gte
-P row_count=164572800 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 18) T2F3-4 , T3F3 , T3F5 -> -r 90 (= ~360)
- >-
latte run --tag latte-main-18 --duration 60m --request-timeout 30 --retry-interval '2s,10s'
--sampling 5s --threads 30 --concurrency 120 --rate 360
--function t2__count,t2__get,t3__count,t3__get_all
-P row_count=164572800 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 19) T4F12/t4__insert_delete_by_1 -> -r 1k (2x500), deletion scenario 1 - by 1
- >-
latte run --tag latte-main-19 --duration 60m --request-timeout 30 --retry-interval '1s,5s'
--sampling 5s --threads 30 --concurrency 90 --rate 500
--function t4__insert_delete_by_1 -P row_count=1000000 -P offset=164572800 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
# 20) T4F13/t4__insert_delete_by_many -> -r 1k, deletion scenario 2 - by many
- >-
latte run --tag latte-main-20 --duration 60m --request-timeout 30 --retry-interval '1s,5s'
--sampling 5s --threads 30 --concurrency 90 --rate 1000
--function t4__insert_delete_by_many -P row_count=1000000 -P offset=165572800 -P print_applied_func_names=2
scylla-qa-internal/custom_d1/workload1/latte/custom_d1_workload1.rn
13 changes: 13 additions & 0 deletions ...es/performance/latte-perf-regression-latency-steady-state-custom-d1-workload1.jenkinsfile
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,13 @@
#!groovy

// trick from https://github.com/jenkinsci/workflow-cps-global-lib-plugin/pull/43
def lib = library identifier: 'sct@snapshot', retriever: legacySCM(scm)

perfRegressionParallelPipeline(
backend: "gce",
region: "us-east1",
test_name: "performance_regression_test.PerformanceRegressionTest",
test_config: '''["configurations/performance/latte-perf-regression-latency-steady-state-custom-d1-workload1.yaml", "configurations/tablets_disabled.yaml", "configurations/disable_kms.yaml"]''',
sub_tests: ["test_latency_steady_state"],
email_recipients: '', // disable sending email
)
1 change: 1 addition & 0 deletions vars/perfRegressionParallelPipeline.groovy
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -265,6 +265,7 @@ def call(Map pipelineParams) {
loadEnvFromString(params.extra_environment_variables)
dir('scylla-cluster-tests') {
checkout scm
checkoutQaInternal(params)
}
dockerLogin(params)
}
Expand Down