Update tileer load test script

[jwalgran]

Update the tile request load testing scripts to

1. Allow running multiple sets of requests in parallel
2. Break the cache for each set of parallel requests to make the load as aggressive as possible

Connects #1776

Demo

> VU_MULTIPLIER=10 \
> CACHE_KEY_SUFFIX=$(echo $(date) | sha1sum | cut -c1-8) \
> docker-compose -f docker-compose.yml run --rm \
>   k6 run /scripts/zoom_rio_de_janerio_with_contributor_filter.js

          /\      |‾‾| /‾‾/   /‾‾/
     /\  /  \     |  |/  /   /  /
    /  \/    \    |     (   /   ‾‾\
   /          \   |  |\  \ |  (‾)  |
  / __________ \  |__| \__\ \_____/ .io

  execution: local
     script: /scripts/zoom_rio_de_janerio_with_contributor_filter.js
     output: -

  scenarios: (100.00%) 1 scenario, 100 max VUs, 10m30s max duration (incl. graceful stop):
           * default: 100 iterations shared among 100 VUs (maxDuration: 10m0s, gracefulStop: 30s)


running (00m16.3s), 000/100 VUs, 100 complete and 0 interrupted iterations
default ✓ [======================================] 100 VUs  00m16.3s/10m0s  100/100 shared iters

     ✓ is status 200

     checks.........................: 100.00% ✓ 950       ✗ 0
     data_received..................: 3.5 MB  213 kB/s
     data_sent......................: 325 kB  20 kB/s
   ✓ failed requests................: 0.00%   ✓ 0         ✗ 950
     http_req_blocked...............: avg=29.13ms  min=88ns     med=264ns    max=1.02s    p(90)=164.73ms p(95)=196.82ms
     http_req_connecting............: avg=10.99ms  min=0s       med=0s       max=161.78ms p(90)=82.33ms  p(95)=98.8ms
     http_req_duration..............: avg=2.24s    min=196.15ms med=2.38s    max=3.93s    p(90)=3.35s    p(95)=3.59s
       { expected_response:true }...: avg=2.24s    min=196.15ms med=2.38s    max=3.93s    p(90)=3.35s    p(95)=3.59s
     http_req_failed................: 0.00%   ✓ 0         ✗ 950
     http_req_receiving.............: avg=101.85µs min=20.42µs  med=83.64µs  max=2.18ms   p(90)=164.75µs p(95)=208.28µs
     http_req_sending...............: avg=141.91µs min=23.51µs  med=102.45µs max=2.56ms   p(90)=262.91µs p(95)=360.88µs
     http_req_tls_handshaking.......: avg=10.55ms  min=0s       med=0s       max=197.98ms p(90)=72.39ms  p(95)=97.47ms
     http_req_waiting...............: avg=2.24s    min=195.6ms  med=2.38s    max=3.93s    p(90)=3.35s    p(95)=3.59s
     http_reqs......................: 950     58.341884/s
     iteration_duration.............: avg=10.59s   min=4.67s    med=9.99s    max=16.24s   p(90)=15.25s   p(95)=16.1s
     iterations.....................: 100     6.141251/s
     vus............................: 9       min=9       max=100
     vus_max........................: 100     min=100     max=100

Testing Instructions

• cd into load-tests
• Run the example from the README and verify that it completes

VU_MULTIPLIER=10 \
CACHE_KEY_SUFFIX=$(echo $(date) | sha1sum | cut -c1-8) \
docker-compose -f docker-compose.yml run --rm \
  k6 run /scripts/zoom_rio_de_janerio_with_contributor_filter.js

• Repeat the test run with VU_MULTIPLIER=20. You should see higher latency and some failed responses
• View the staging CloudWatch dashboard to verify the increases in request, latency, and DB CPU.

Checklist

• fixup! commits have been squashed
• CI passes after rebase
• CHANGELOG.md updated with summary of features or fixes, following Keep a Changelog guidelines

[rajadain]

Taking a look

[rajadain]

+1 tested. Tried with 10, 20, and 30 virtual users. Got 0%, 2%, and 20% failed requests respectively. This is a great setup, hopefully can inform future scaling decisions.

[rajadain]

It took me a while to figure out where __VU was defined: https://k6.io/docs/using-k6/execution-context-variables/#_vu-and-_iter-discouraged. Seemingly discouraged by the docs, but works as expected here.

[jwalgran]

I noticed the discouragement, too, but continued using __VU since it was already used successfully elsewhere in the script.

[rajadain]

Also interesting seeing the staging load metrics:

For higher VUs, the ramp up was much steeper.

[jwalgran]

Thanks for the review.