Why does iostat and sar -d sometimes output incorrect results?
Issue
- On machines with heavy io loads, the output of sar -d is faulty, displaying some very large numbers.
- Using nr_requests for testing with high values (50,000+), iostat displays incorrect results fairly frequently.
- Example of "bad" sar -d output:
09:20:01 PM DEV tps rd_sec/s wr_sec/s avgrq-sz avgqu-sz await svctm %util
09:25:01 PM sdb1 0.52 0.03 5.52 10.60 0.00 1.27 1.02 0.05
09:30:01 PM sdb1 0.54 0.19 5.57 10.73 0.00 1.74 1.30 0.07
09:35:01 PM sdb1 33.64 259.94 47.26 9.13 0.04 1.17 0.29 0.96
09:40:01 PM sdb1 2202.02 6257.93 24939.09 14.17 2078.79 943.69 0.30 66.15
09:45:01 PM sdb1 3051.57 0.27 36532.41 11.97 1989.30 638.13 0.28 86.88
09:50:01 PM sdb1 2986.02 0.05 34704.99 11.62 3355.91 1137.22 0.28 84.49
09:55:01 PM sdb1 3284.08 0.21 32884.03 10.01 4556.98 1386.29 0.27 88.31
10:00:01 PM sdb1 3260.33 0.19 31497.61 9.66 61474802782539.38 18855397092136.58 0.27 87.60
10:05:01 PM sdb1 3373.57 0.08 32028.26 9.49 6069.53 1799.10 0.27 89.86
Environment
- Red Hat Enterprise Linux 5
- Sysstat 7.0.2 (reproduced in upstream version 10.0.2)
- Red Hat Enterprise Linux 6
- Sysstat version below 9.0.4-20
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.