Hide Forgot
Description of problem: I have noticed for a while that iostat on RHEL 5 seems to show artificially low values on the first iteration if "interval is set", or, on its only iteration if invoked without the "interval" parameter. I noticed recently that on a Debian system (with sysstat v 8.1.2), the same problem doesn't seem to happen. If this is a bug that has been fixed rather than an inherent limitation of the RHEL version of sysstat, please fix it. It's not super obvious when i/o is low, but it's very noticeable / reproducible when the system is more heavily loaded in a way that makes me suspect it's not a simple matter iowait fluctuating. For example, even when the second and later iterations of iowait report iowait % in the 40-60% range, the first reading will often be under 2%. Version-Release number of selected component (if applicable): 7.0.2-11.el5 How reproducible: Very, however, it's easier to notice when the system is under heavy i/o or heavy load. Steps to Reproduce: 1. Run iostat N 2. First %iowait (and possibly other, like %user) value will be artificially low Actual results: Expected results: Additional info:
Running iostat over and over gives me the same %iowait, %user, etc. values consistently: water-griffen:$ iostat Linux 2.6.18-274.7.1.el5PAE (water-griffen.caltech.edu) 02/16/2012 avg-cpu: %user %nice %system %iowait %steal %idle 1.28 0.00 1.43 3.55 0.00 93.74 Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn sda 38.47 4.20 536.10 13127193 1675426004 sda1 0.00 0.00 0.00 424 0 sda2 0.00 0.00 0.00 500 0 sda3 0.00 0.00 0.00 3282 6948 sda4 0.00 0.00 0.00 6 0 sda5 38.47 4.20 536.10 13112602 1675418496 sda6 0.00 0.00 0.00 1459 560 water-griffen:$ iostat Linux 2.6.18-274.7.1.el5PAE (water-griffen.caltech.edu) 02/16/2012 avg-cpu: %user %nice %system %iowait %steal %idle 1.28 0.00 1.43 3.55 0.00 93.74 Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn sda 38.47 4.20 536.10 13127193 1675426540 sda1 0.00 0.00 0.00 424 0 sda2 0.00 0.00 0.00 500 0 sda3 0.00 0.00 0.00 3282 6948 sda4 0.00 0.00 0.00 6 0 sda5 38.47 4.20 536.10 13112602 1675419032 sda6 0.00 0.00 0.00 1459 560 water-griffen:$ iostat Linux 2.6.18-274.7.1.el5PAE (water-griffen.caltech.edu) 02/16/2012 avg-cpu: %user %nice %system %iowait %steal %idle 1.28 0.00 1.43 3.55 0.00 93.74 Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn sda 38.47 4.20 536.10 13127193 1675426540 sda1 0.00 0.00 0.00 424 0 sda2 0.00 0.00 0.00 500 0 sda3 0.00 0.00 0.00 3282 6948 sda4 0.00 0.00 0.00 6 0 sda5 38.47 4.20 536.10 13112602 1675419032 sda6 0.00 0.00 0.00 1459 560
And, running iostat 5 shows about that same value on the first iteration, and the (apparently) correct value on subsequent iterations: water-griffen:$ iostat 5 Linux 2.6.18-274.7.1.el5PAE (water-griffen.caltech.edu) 02/16/2012 avg-cpu: %user %nice %system %iowait %steal %idle 1.27 0.00 1.43 3.55 0.00 93.74 Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn sda 38.46 4.19 536.06 13099411 1675095524 sda1 0.00 0.00 0.00 424 0 sda2 0.00 0.00 0.00 500 0 sda3 0.00 0.00 0.00 3276 6940 sda4 0.00 0.00 0.00 6 0 sda5 38.46 4.19 536.06 13084826 1675088024 sda6 0.00 0.00 0.00 1459 560 avg-cpu: %user %nice %system %iowait %steal %idle 3.16 0.00 4.01 29.28 0.00 63.56 Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn sda 41.60 30.40 604.80 152 3024 sda1 0.00 0.00 0.00 0 0 sda2 0.00 0.00 0.00 0 0 sda3 0.00 0.00 0.00 0 0 sda4 0.00 0.00 0.00 0 0 sda5 41.60 30.40 604.80 152 3024 sda6 0.00 0.00 0.00 0 0 avg-cpu: %user %nice %system %iowait %steal %idle 3.13 0.00 3.41 27.75 0.00 65.71 Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn sda 43.60 0.00 667.20 0 3336 sda1 0.00 0.00 0.00 0 0 sda2 0.00 0.00 0.00 0 0 sda3 0.00 0.00 0.00 0 0 sda4 0.00 0.00 0.00 0 0 sda5 43.60 0.00 667.20 0 3336 sda6 0.00 0.00 0.00 0 0 avg-cpu: %user %nice %system %iowait %steal %idle 4.52 0.00 5.04 27.72 0.00 62.72 Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn sda 64.40 1.60 870.40 8 4352 sda1 0.00 0.00 0.00 0 0 sda2 0.00 0.00 0.00 0 0 sda3 0.00 0.00 0.00 0 0 sda4 0.00 0.00 0.00 0 0 sda5 64.40 1.60 870.40 8 4352 sda6 0.00 0.00 0.00 0 0 avg-cpu: %user %nice %system %iowait %steal %idle 3.96 0.00 4.19 26.41 0.00 65.44 Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn sda 102.00 0.00 1330.14 0 6664 sda1 0.00 0.00 0.00 0 0 sda2 0.00 0.00 0.00 0 0 sda3 0.00 0.00 0.00 0 0 sda4 0.00 0.00 0.00 0 0 sda5 102.00 0.00 1330.14 0 6664 sda6 0.00 0.00 0.00 0 0 avg-cpu: %user %nice %system %iowait %steal %idle 3.98 0.00 3.85 29.69 0.00 62.47 Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn sda 57.40 0.00 856.00 0 4280 sda1 0.00 0.00 0.00 0 0 sda2 0.00 0.00 0.00 0 0 sda3 0.00 0.00 0.00 0 0 sda4 0.00 0.00 0.00 0 0 sda5 57.40 0.00 856.00 0 4280 sda6 0.00 0.00 0.00 0 0 avg-cpu: %user %nice %system %iowait %steal %idle 6.23 0.00 6.54 28.80 0.00 58.43 Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn sda 281.60 3249.60 3105.60 16248 15528 sda1 0.00 0.00 0.00 0 0 sda2 0.00 0.00 0.00 0 0 sda3 0.00 0.00 0.00 0 0 sda4 0.00 0.00 0.00 0 0 sda5 281.60 3249.60 3105.60 16248 15528 sda6 0.00 0.00 0.00 0 0
I also created Red Hat case #00600185
The behavior is actually not as different as I thought with sysstat 8 on Debian (the iostat % seems to be higher rather than lower, but the other behavior is still the same). The behavior still strikes me as unusual.
Ahah. I looked in TFM before, but not in the right place... near the top: "The first report generated by the iostat command provides statistics concerning the time since the system was booted." You can close this.