Hi Pieter, On Mon, Jun 10, 2019 at 04:06:13PM +0200, PiBa-NL wrote: > Things certainly look better again now regarding this issue.
Ah cool! > Running the test repeatedly, and manually looking over the results its > pretty much as good as it was before. There seems to be a 1 ms increase in > the check-duration, but maybe this is because of the moved initialization > which on startup delays the first test a millisecond or something? It should not. At this point I think it can be anything including measurement noise or even thread assigment on startup! > Below some test results that are based on manual observation and some in my > head filtering of the console output.. (mistakes included ;) ) > repeat 10 ./vt -v ./work/haproxy-*/reg-tests/checks/tls_health_checks.vtc | > grep Layer7 | grep OK | grep WARNING > Commit-ID , min-max time for +-95% check durations , comment > e4d7c9d , 6 - 9 ms , all tests pass (1 tests out of +- a hundred showed > 29ms , none below 6ms and almost half of them show 7ms) Great! > 6ec902a , 11 - 150 ms , of the 12 tests that passed That's quite a difference indeed. > e186161 , 5 - 8 ms , all tests pass (1 test used 15 ms, more than half > the tests show 5ms check duration the majority of the remainder show 6ms) OK! > I'm not sure if this deserves further investigation at the moment, i think > it does not. Thanks for spending your weekend on this :) that wasn't my > intention. Oh don't worry, you know I'm a low-level guy, just give me a problem to solve with a few bits available only and I can spend countless hours on it! Others entertain themselves playing games, for me this is a game :-) Thanks a lot for testing, at least we know there isn't another strange thing hidden behind. Cheers, Willy