Improve analysing mechanism of HDR file

[juliayakovlev]

We can improve analysing mechanism of HDR file by running build_histograms_summary_with_interval with shorter interval (in the test we perform summary for 600 sec interval), like:

hdr_file = "/home/juliayakovlev/Downloads/loader-set-8af5fdae/"

start_time = 1743691356.5128493
end_time = 1743696279.4485803

interval_sec = 30
res = make_hdrhistogram_summary_by_interval(hdr_tags=["WRITE-st"], path=hdr_file, stress_operation="WRITE", start_time=start_time, end_time=end_time, interval=interval_sec)
p99 = {}
for num, read in enumerate(res):
    dt = datetime.utcfromtimestamp(start_time + num * interval_sec).strftime('%Y-%m-%d %H:%M:%S')
    p99[dt] = read['WRITE']['percentile_99']
print(p99)

Possible result:

'2025-04-03 14:42:36': 62.0, 
'2025-04-03 14:43:06': 72.48, 
'2025-04-03 14:43:36': 77.66, 
'2025-04-03 14:44:06': 79.63, 
'2025-04-03 14:44:36': 65.7, 
'2025-04-03 14:45:06': 63.64, 
'2025-04-03 14:45:36': 82.97, 
'2025-04-03 14:46:06': 113.51, 

According to result we can find in the cluster logs what happened in the time of high latency

[fruch]

@juliayakovlev is that the helper tool we were talking about ?