Setup:
I am currently running my k6 test in a container(4 cpu, 16Gig mem) which I ran by building the xk6 image using the Dockerfile mentioned xk6-output-influxdb/Dockerfile at main · grafana/xk6-output-influxdb · GitHub
I am collecting and sending the metrics to InfluxDb using xk6 and then plotting a graph of requests/sec in grafana.
Test file
I am running the following test against a server
import http from 'k6/http';
import { check, sleep } from 'k6';
export const options = {
scenarios: {
cs: {
executor: 'constant-vus',
vus: 2,
duration: '60s',
// stages: [
// { duration: '10s', target: 10 },
// { duration: '50s', target: 10 },
// { duration: '10s', target: 0 },
// ]
}
}
};
export default function () {
const url = 'http://ai-gateway-influx:5052/v2/code/completions'; // Replace with your API endpoint
const payload = JSON.stringify({
"project_path": "string",
"project_id": 0,
"current_file": {
"file_name": "test",
"language_identifier": "string",
"content_above_cursor": "func hello_world(){\n\t",
"content_below_cursor": "\n}"
},
"stream": true,
"choices_count": 0,
"context": [],
"prompt_id": "code_suggestions/generations",
"prompt_version": 2
});
const params = {
headers: {
'Content-Type': 'application/json',
},
};
const res = http.post(url, payload, params);
check(res, {
'status is 200': (r) => r.status === 200,
'response time < 500ms': (r) => r.timings.duration < 500,
});
const utcTimestamp = new Date().toISOString();
const message = `{"pe_iteration":"${__ITER}", "pe_http_method": "${res.request.method}", "pe_http_url": "${res.request.url}", "pe_http_status": "${res.status}", "pe_response_duration": "${res.timings.duration}ms", "pe_utc_timestamp": "${utcTimestamp}"}`
console.log(JSON.parse(message))
sleep(1);
}
Actual Results: When I run the tests, I can see that intermittently, the number of requests go down and then come back up again.
Expected Result: I would expect the load that is being generated by k6 to be constant.
You can see in the below logs that there is no log entry for 2025-01-21T01:15:56 but there is one for 2025-01-21T01:15:55 and another for 2025-01-21T01:15:57
You can also see this drop which is pretty obvious in the Grafana dashboard. In this case it dropped to 0 from 2rps but when I tried for more VUs like 20 it used to drop to say 7/8/other lower value and then back to 20.
In case you are wondering the query I used for grafana
from(bucket: "test")
|> range(start: v.timeRangeStart, stop: v.timeRangeStop)
|> filter(fn: (r) => r._measurement == "http_reqs")
|> aggregateWindow(every: 1s, fn: count)
I also tried increasing the container capacity by increasing the CPU and memory but to no avail. I can still see requests dropping in between during the duration of the test.
Can you help me figure out why this is the case?
