From f5999cb21a45e57883858da03b95b90905c01ab0 Mon Sep 17 00:00:00 2001
From: Bo Xu <shifucun@users.noreply.github.com>
Date: Thu, 23 May 2024 05:17:41 +0000
Subject: [PATCH] [NL Load Testing] Support various instance and fix a bug when
specify total requests (#4267)
---
tools/nl/loadtest/README.md | 17 ++++++++++++-----
tools/nl/loadtest/load.py | 9 +++++----
2 files changed, 17 insertions(+), 9 deletions(-)
diff --git a/tools/nl/loadtest/README.md b/tools/nl/loadtest/README.md
index 7dbb49066d..5668d87ddc 100644
--- a/tools/nl/loadtest/README.md
+++ b/tools/nl/loadtest/README.md
@@ -1,10 +1,12 @@
+# NodeJS Server Load Testing
+
## NL SVG Latency tester
A few tens of queries against the NodeJS NL server (`nodejs/query`) in series and tracks times.
Run it as:
-```
+```bash
./run_latencies.sh
```
@@ -14,14 +16,19 @@ Runs specified number of queries against the NodeJS NL server (`nodejs/query`),
Run it as:
-```
-./run_load.sh --total_requests=<total queries to send> --parallel_requests=<number of queries to send at a time> --apikey=<apikey for bard instance>
+```bash
+./run_load.sh \
+ --instance=<bard|dev> \
+ --total_requests=<total queries to send> \
+ --parallel_requests=<number of queries to send at a time> \
+ --apikey=<apikey for bard instance>
```
Notes about the arguments:
+
- If total requests is not set, this tool will run through all the queries in queryset.csv once
- If parallel requests is not set, this tool will run requests one after another in series
- With the current setup, 20000 total requests and 150 parallel requests will run for about 7 mins and achieve a qps of around 45
- For the apikey, you can go to the [Bard Apigee Console](https://pantheon.corp.google.com/apigee/apps?mods=-monitoring_api_staging&project=datcom-bard) and either:
- 1. Use an existing key
- 2. Create your own key by [creating an "API Product", "App", and "Developer"](https://cloud.google.com/apigee/docs/api-platform/security/api-keys).
\ No newline at end of file
+ 1. Use an existing key
+ 2. Create your own key by [creating an "API Product", "App", and "Developer"](https://cloud.google.com/apigee/docs/api-platform/security/api-keys).
diff --git a/tools/nl/loadtest/load.py b/tools/nl/loadtest/load.py
index 3dfddcad23..07d7657a2c 100644
--- a/tools/nl/loadtest/load.py
+++ b/tools/nl/loadtest/load.py
@@ -22,6 +22,7 @@
FLAGS = flags.FLAGS
+flags.DEFINE_string('instance', 'bard', 'The instance name of the website')
flags.DEFINE_integer('parallel_requests', 1,
'Number of requests to run in parallel')
flags.DEFINE_integer('total_requests', None, 'Total number of requests to run')
@@ -29,7 +30,7 @@
_OUTPUT_FILE = 'load_results.json'
_QUERY_FILE = 'queryset.csv'
-_URL = 'https://bard.datacommons.org/nodejs/query?apikey={apikey}&q={query}'
+_URL = 'https://{instance}.datacommons.org/nodejs/query?apikey={apikey}&q={query}'
_RESULT_KEY_CODE = 'code'
_RESULT_KEY_ERROR_CONTENT = 'errorContent'
_RESULT_KEY_NL_TIME = 'nlTime'
@@ -43,8 +44,8 @@ def _run_query(query):
# Make the API request
print(f'Running query: {query}')
try:
- response = session.get(_URL.format(apikey=FLAGS.apikey, query=query),
- timeout=None)
+ url = _URL.format(instance=FLAGS.instance, apikey=FLAGS.apikey, query=query)
+ response = session.get(url, timeout=None)
if response.status_code == 200:
debug_timing = response.json().get('debug', {}).get('timing', {})
return {
@@ -129,7 +130,7 @@ def run_load(total_requests, parallel_requests):
# If number of total requests is specified, generate a list of queries that is
# total requests long
if total_requests:
- while len(query_list_to_run) < total_requests or len(query_list):
+ while len(query_list_to_run) < total_requests and len(query_list):
num_missing = total_requests - len(query_list_to_run)
query_list_to_run += query_list[0:min(num_missing, len(query_list))]
else: