Vmstat specifics

[John Oberlander]

Im running a rp9000, 10CPU, 18Gram, storage works eva for a Universe database specific to collections. The question i have is what does the "blocked" field on vmstat really mean? We has major slowness last night on some jobs and the blocked field hit 79 on my custom cacti graph that i built during the same time of the slowness issues. The reports during the night are 90% reads from the eva and 10% writes. We dont seem to be using no where near the max throughput on the eva and the data is being blocked inside somewhere. Our load is never over 1, buffercache hits are 90% or higher. GPM now shows 12% blocked by system, 6% other, and 0 on everything else. Where do i go from here? What does blocked on system mean? Our cpu's arn't being touched.

Thanks,
John

[Tony Scully_2 · Accepted solution]

Can you post swapinfo -tam

Tony

[John Oberlander]

universe@/: swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4096 0 4096 0% 0 - 1 /dev/vg00/lvol2
dev 6144 0 6144 0% 0 - 0 /dev/vg00/lvol9
reserve - 2407 -2407
memory 13806 6480 7326 47%
total 24046 8887 15159 37% - 0 -
universe@/:

[Tony Scully_2]

Also, running sar what sort of values do you see in the %wio field?

T

[John Oberlander]

universe@/: sar 5 5

HP-UX universe B.11.11 U 9000/800 03/25/05

16:01:51 %usr %sys %wio %idle
16:01:56 8 20 17 56
16:02:01 7 14 11 68
16:02:06 5 20 13 61
16:02:11 8 17 7 68
16:02:16 11 11 9 69

Average 8 16 11 64

[John Oberlander]

That was right now...here is from last night when the issues were happening

04:00:00 22 14 62 2
04:10:00 22 12 65 1
04:20:00 23 13 63 1
04:30:00 21 12 66 1
04:40:00 22 13 64 1
04:50:00 21 13 65 1
05:00:00 20 13 66 1
05:10:00 18 10 71 1
05:20:00 18 10 69 2
05:30:00 18 11 69 2
05:40:00 22 14 62 1
05:50:00 22 15 61 2
06:00:00 21 16 60 3
06:10:00 29 20 49 2
06:20:00 27 20 51 2
06:30:01 20 14 63 3
06:40:00 22 20 55 2
06:50:00 29 21 48 1
07:00:00 25 18 55 2
07:10:00 23 23 52 2
07:20:00 30 32 37 1
07:30:00 27 26 46 2
07:40:00 30 28 41 1
07:50:00 24 28 47 1
08:00:00 18 23 58 1

[Tony Scully_2]

John,

Thats a lot of I/O wait. I don't know the EVA, but I would run some diags on that (I assume it has arm* type commands) -- could it have been rebuilding a failed disk onto a hot spare?

Need a storage works expert here.

Sorry for not being much assistance.

T

[John Oberlander]

Thanks, ill do some investigation on that side.