Hello,
I have a custom workload as below... Post the run of this custom workload, it always complains about one particular workload definition for skew being out of sync. Can somebody please suggest how to fix this. I tried changing warmup time and elapsed time to as large as 3600 (1hr), but that also didnt solve -
wd=custom_wl_wd_RRH,sd=sd*,seekpct=100,rhpct=100,rdpct=100,skew=34,range=(0,100),xfersize=30k
wd=custom_wl_wd_RRM,sd=sd*,seekpct=100,rdpct=100,skew=24,range=(0,100),xfersize=36k
wd=custom_wl_wd_RWM,sd=sd*,seekpct=100,rdpct=0,skew=16,range=(0,100),xfersize=24k
wd=custom_wl_wd_SR,sd=sd*,seekpct=0,rdpct=100,skew=13,range=(0,100),xfersize=40k
wd=custom_wl_wd_SW,sd=sd*,seekpct=0,rdpct=0,skew=8,range=(0,100),xfersize=32k
wd=custom_wl_wd_RWH,sd=sd*,seekpct=100,whpct=100,rdpct=0,skew=5,range=(0,100),xfersize=24k
rd=custom_wl,wd=custom_wl_wd*,iorate=(max),interval=10,elapsed=600,threads=16,forhitarea=(20m),warmup=600
Post the completion of run, it always complains -
04:55:48.284 ************************************************************************************************************
04:55:48.284 * Observed Workload skew for wd=custom_wl_wd_SR delta greater than 1% (9.54). See skew.html for details *
04:55:48.296 ************************************************************************************************************
04:55:48.297 *
04:55:48.297 *
04:55:48.298 *************************************************************************************************************************************
04:55:48.298 * Possible reasons for workload skew being out of sync: *
04:55:48.298 * *
04:55:48.298 * - Elapsed time too short. *
04:55:48.298 * Each SD should have done at least 2000 i/o's. Adding 'warmup=nn' may help this. *
04:55:48.298 * For SD concatenation this is the ONLY acceptable reason. Contact me at the Oracle Vdbench Forum if the differences are too big. *
04:55:48.298 * *
04:55:48.299 * - A mix of seekpct=seq and random workloads. *
04:55:48.299 * Sequential workloads may run on only ONE slave, but random workloads *
04:55:48.299 * may run on multiple slaves. Workload skew depends on ALL workloads *
04:55:48.299 * running on all slaves. Force slave count to just one. (hd=default,jvms=1) *
04:55:48.299 * *
04:55:48.299 * - Not enough threads available to guarantee that all workloads can run *
04:55:48.299 * on all slaves. *
04:55:48.299 * *
04:55:48.299 * - There may be more, can't think (of any) right now. :-) *
04:55:48.299 * *
04:55:48.299 * You may use 'abort_failed_skew=nn' for Vdbench to abort after skew failures. See documentation. *
04:55:48.299 * *
04:55:48.299 *************************************************************************************************************************************
And skew.html gives the below -
04:55:48.281 Calculated versus requested workload skew. (Delta only shown if > 0.10% absolute)
04:55:48.281 Note that for an Uncontrolled MAX run workload skew is irrelevant.
04:55:48.281
04:55:48.281 i/o MB/sec bytes read resp read write resp resp queue skew skew skew
04:55:48.281 WD: rate 1024**2 i/o pct time resp resp max stddev depth requested observed delta
04:55:48.282 custom_wl_wd_RRH 37993.23 1113.08 30720 100.00 2.821 2.821 0.000 226.635 1.557 107.2 34.00% 40.63% -6.63%
04:55:48.282 custom_wl_wd_RRM 26820.21 942.90 36864 100.00 1.485 1.485 0.000 33.685 0.914 39.8 24.00% 28.68% -4.68%
04:55:48.283 custom_wl_wd_RWM 17878.32 419.02 24576 0.00 1.292 0.000 1.292 22.674 0.579 23.1 16.00% 19.12% -3.12%
04:55:48.283 custom_wl_wd_SR 3235.30 126.38 40960 100.00 2.097 2.097 0.000 22.969 1.730 6.8 13.00% 3.46% 9.54%
04:55:48.283 custom_wl_wd_SW 1992.07 62.25 32768 0.00 1.305 0.000 1.305 12.682 0.577 2.6 8.00% 2.13% 5.87%
04:55:48.283 custom_wl_wd_RWH 5589.42 131.00 24576 0.00 1.351 0.000 1.351 19.552 0.637 7.6 5.00% 5.98% -0.98%
04:55:48.283 Total 93508.55 2794.64 31338 72.77 2.000 2.260 1.306 226.635 1.380 187.0 100.00% 100.00%