1st set of stress testing

[Shubham8287]

No description provided.

[Shubham8287]

Going through this - https://llvm.org/docs/Benchmarking.html

[Shubham8287]

and this - https://www.cis.upenn.edu/~cis501/previous/fall2017/papers/producing-wrong-data.pdf

[Shubham8287]

Checking node exporter

[RiteshKSingh1709]

what are the result of above two btw @Shubham8287 ? you can also put the findings help others to understand and easier for you as well to compare.

[Shubham8287]

Tried node exporter + prometheus + grafana for systems metrcics, the system is overkilling for us. Realized all such tools rely on basic linux commands like top, iostat, etc but anyway got good exposed to collect metrics of large system.

[Shubham8287]

I was doing stress testing, started with pi but memory card is bottleneck there CPU spending 99% time idle.
for go server, it is not accepting more than 1000 concurrent requests, failing with error "too many open files".

tried healthcheck benchmark with 'ab', found something strange.
"ab -n 200000 -c 1000 http://localhost:8080/isAlive", returns 116ms median whereas " ab -n 200000 -c 1 http://localhost:8080/isAlive", returns 1ms median. Ideally response time should decrease if making request in parallel, might be due to caching. need to check.

[Shubham8287]

this thread fixed open files error - https://askubuntu.com/questions/1182021/too-many-open-files