Hi All,
If this is posted in the wrong section of the forum, please direct me to the correct section.
This is RHEL 6.3, kernel version 2.6.32-279.14.1.el6.x86_64.
I'm seeing some "weirdness" pertaining to one partition on one device from /proc/diskstats.
Specifically, I'm seeing a very large number in field 9 (# of I/Os currently in progress) for one partition of one device (I right justified the numbers to make them more human readable):
Code:
17:09:23 29-JUL-13
…
120 128 emcpoweri 23027652 1873 749143373 29134240 12218284 91910 372823337 5777842 0 25840679 34889691
120 129 emcpoweri1 23027394 1860 749141205 29134211 12218284 91910 372823337 5777842 0 25840651 34889662
120 112 emcpowerh 11437540 1206 655430701 28168210 11389069 88581 363132716 5402079 0 24553665 33555888
120 113 emcpowerh1 11437282 1193 655428533 28168185 11389069 88581 363132716 5402079 0 24553647 33555870
120 160 emcpowerk 66307936 1786 750055270 48079658 14862294 88253 1265181343 10406951 0 33308680 58436025
120 161 emcpowerk1 66307222 1718 750049014 48079559 14862287 88234 1265181135 10406928 4294967295 705106587 3619973254
…
17:34:17 29-JUL-13
…
120 128 emcpoweri 23201721 1932 751462747 29321438 12324220 92848 374543849 5821521 0 26056823 35120379
120 129 emcpoweri1 23201463 1919 751460579 29321409 12324220 92848 374543849 5821521 0 26056795 35120350
120 112 emcpowerh 11519074 1225 657069064 28344914 11487551 89491 364723951 5441962 0 24756914 33772379
120 113 emcpowerh1 11518816 1212 657066896 28344889 11487551 89491 364723951 5441962 0 24756896 33772361
120 160 emcpowerk 66864771 1839 754835274 48569086 14993163 89077 1267236690 10478880 0 33611662 58996968
120 161 emcpowerk1 66864057 1771 754829018 48568987 14993156 89058 1267236482 10478857 4294967295 710004774 3615358538
…
(We are using EMC PowerPath for data path management and the host was recently rebooted.)
I expect all of these LUNs (emcpoweri, emcpowerh and emcpowerk) to have relatively the same stats (they are 3 parts of a very large Oracle ASM diskgroup containing a number of LUNs which is "load balanced").
I would also expect to see the values for the LUNs and partitions be very close to each other since this is the only partition on each LUN (taking into account the minor inaccuracies due to lack of locks).
However, I do not expect to see the value of field 9 (as well as it's dependant fields 10 and 11) be this large [ever].
I also do not expect to see a 0 for the LUN and 4294967295 for the single partition on that same LUN.
I know that value (4294967295) is the max for a 32-bit number, however, I'd really like to understand:
1. Why is that number so disproportionately large as opposed to the others (0 for other LUNs as well as partitions)?
2. Why does it stay that way?
Thanks,
Rich
