[test][CI] benchmark under both managers are down

[mocchira]

(IIRC, this kind of benchmarks is not conducted regularly.) With the fix for https://github.com/leo-project/leofs/issues/642, As leo_gateway(s) have to ping both managers every time in case handling object operations via s3 while both are down, performance degradation may happen so at least we have to benchmark to confirm there is negligible performance degradation also it would be great to get this benchmark suite into the CI pipeline.

[windkit]

As a first test, I tested the scenario when Manager are down The performance is very close to normal case

Report at: https://github.com/leo-project/notes/tree/master/leofs/benchmark/leofs/1.3/1.3.3/front/20170331_image_f4m_load_read_manager_30min_1

Load

Normal

Managers Down

Read

Normal

Managers Down

[mocchira]

@windkit thanks.

looks good. a little bit throughput decreasing is within my expectation.

then would the next step be to get this suite into concourse CI pipeline?