Assertion failure while running benchmark about wrk2

[kklis]

A friend of mine suggested that the problem may occur when NTP time correction happens while running wrk2, and the system clock turns back. The possibility of such event seems to be much more higher in docker than in a regular system.

[ykameshrao]

I am hitting into this one. Any solutions yet? Or is it way obvious!

[kklis]

I've had similar problem today. Running in Ubuntu container on Mac OS X High Sierra (Docker version 18.06.1-ce-mac73 (26764)):

root@ef5b3a1d8be9:~/wrk2# ./wrk2 --version
wrk 4.0.0 [epoll] Copyright (C) 2012 Will Glozer

root@ef5b3a1d8be9:~/wrk2# ./wrk2 -t1 -c1 -d300m -R10 http://...
Running 300m test @ http://...
  1 threads and 1 connections
  Thread calibration: mean lat.: 40.739ms, rate sampling interval: 90ms


 ----------

We are about to crash and die (recoridng a negative #)This wil never ever ever happen...But when it does. The following information will help in debuggingresponse_complete:
  expected_latency_timing = -1340
  now = 1537524361245120
  expected_latency_start = 1537524361246460
  c->thread_start = 1537524299946460
  c->complete = 614
  throughput = 1e-05
  latest_should_send_time = 1537524361247900
  latest_expected_start = 1537524361247848
  latest_connect = 1537524299946643
  latest_write = 1537524361247907
  next expected_latency_start = 1537524361346460
wrk2: src/hdr_histogram.c:54: counts_index: Assertion `bucket_index < h->bucket_count' failed.
Aborted

Re-running works ok.

[dabos]

-R should be equal or higher than -c