On Mon, 02/10 13:32, Benoît Canet wrote: > Le Friday 07 Feb 2014 à 18:43:33 (+0800), Fam Zheng a écrit : > > + nsec_per_sec = 1000000000 > > + > > + # With current interface we can't set burst to 0, they will be > > override > > + # to bps / 10 and iops / 10. Setting to 1 workarounds this and > > gets to > > + # a close effect as no burst, which is good enough for us > > + limits['bps_max'] = 1 > > + limits['iops_max'] = 1 > Maybe the code should also set *rd_max = 1 and *wr_max= 1 to make the *_rd_* > and > *_wr_* tests more accurate.
The code doesn't allow setting bps_rd_max together with bps_max. I think bps_max has enough effective on burst, so it's not strictly necessary. Fam
