On Tue, Mar 20, 2018 at 08:59:21AM +0100, Dominique Martinet wrote: > Hi, > > Girjesh Rajoria wrote on Mon, Mar 19, 2018 at 10:47:09PM +0530: > >> + tail -21 ../dbenchTestLog.txt > >> > >> Operation Count AvgLat MaxLat > >> -------------------------------------------------- > >> Deltree 102 9.799 27.590 > >> Flush 284316 1.637 203.259 > >> Close 2979801 0.007 0.330 > >> LockX 13208 0.007 0.079 > >> Mkdir 51 0.011 0.059 > >> Rename 171774 0.073 0.463 > >> ReadX 6358865 0.010 38.319 > >> WriteX 2022375 0.048 40.888 > >> Unlink 819204 0.090 38.363 > >> UnlockX 13208 0.006 0.063 > >> FIND_FIRST 1421549 0.044 38.320 > >> SET_FILE_INFORMATION 330438 0.024 0.310 > >> QUERY_FILE_INFORMATION 644319 0.004 0.242 > >> QUERY_PATH_INFORMATION 3676827 0.015 40.851 > >> QUERY_FS_INFORMATION 674193 0.010 37.783 > >> NTCreateX 4056560 0.049 122.097 > >> > >> > >> Where are the iozone results from ../ioZoneLog.txt? > > > > iozone suite doesn't give outputs result as dbench. So iozone test checks > > for successful completion and print message of success in the log. In cases > > where the test fails, it'll print error due to which test failed from > > ../ioZoneLog.txt. > > I think it's great to have this kind of dbench stats, and would be > awesome if we can have some raw figures from iozone as well (I think it > can output the results in csv format at least?) > > > jenkins can also take performance metrics from jobs and we could have > graphs of the performance over time if it keeps these metrics a bit > longer than the actual jobs (for example with the performance plugin[1], > but there might be other ways) > > On an individual basis as the tests are on VMs with various loads the > results will probably flicker a bit, but on a whole we should be able to > identify what week(s) introduced slowdowns/speedups after the fact quite > nicely if we can achieve that! :)
Note that the tests in the CentOS CI run on different physical hosts. When a test is started, one or more machines are requested, and the scheduler (called Duffy) just returns a random system. This means that the performance results might differ quite a bit between runs, even for the same change-set. See https://wiki.centos.org/QaWiki/PubHardware for details about the hardware. So except for the performance results, it may be useful to gather some details about the hardware that was used. Niels ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Nfs-ganesha-devel mailing list Nfs-ganesha-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel