sar data - feedback please

[Tom Sekar]

# sar -b 2 25

HP-UX CT45QA B.11.11 U 9000/800 06/28/06

08:14:31 bread/s lread/s %rcache bwrit/s lwrit/s %wcache pread/s pwrit/s
08:14:33 1635 1985 18 80 166 52 0 0
08:14:35 1534 1336 0 15 18 19 0 0
08:14:37 1987 1737 0 15 17 9 4 0
08:14:39 1796 1156 0 41 44 6 0 0
08:14:41 1869 1301 0 40 51 23 0 0
08:14:43 1804 1508 0 150 99 0 0 0
08:14:45 2183 1497 0 99 149 33 0 0
08:14:47 2871 7016 59 37 45 19 0 0
08:14:49 2179 1731 0 58 48 0 0 0
08:14:51 2785 1742 0 110 115 4 0 0
08:14:53 3159 1742 0 23 39 41 0 0
08:14:55 2989 2212 0 115 108 0 0 0
08:14:57 2971 2167 0 60 82 26 0 0
08:14:59 3613 1913 0 60 71 16 0 0
08:15:01 3705 3092 0 160 168 4 0 0
08:15:03 3007 1662 0 52 113 54 0 0
08:15:05 2560 1970 0 58 61 5 0 0
08:15:07 2959 2603 0 56 68 18 0 0
08:15:09 2898 2423 0 53 68 23 0 0
08:15:11 2151 1382 0 42 47 11 0 0
08:15:13 2417 1573 0 114 112 0 0 0
08:15:15 2330 1712 0 72 98 27 0 0
08:15:17 1656 8132 80 227 491 54 0 0
08:15:19 2077 2225 7 105 163 36 0 0
08:15:21 3095 3971 22 108 183 41 0 0

Average 2489 2390 3591766 78 105 26 0 0

The machine's swap usage is always high.

Thanks

Tom

[Darrel Louis]

Hi Tom,

Can you post the following info aswell:

# vmstat 2 20
# swapinfo -tam

Do you have glance installed on the server, to check memory and swap usage.

Darrel

[Steven E. Protter]

Shalom Tom,

vmstat should show heavy paging if swap is stressed.

If so, buy memory.

SEP

[TKeller]

How much physical ram do you have and what does swapinfo show?

Also, what applications is this machine running?

[Michael Steele_2]

what are bufpages, nbuf, dbc_max_pct and dbc_min_pct?

kmtune -l -q dbc_max_pct
kmtune -l -q dbc_min_pct, etc.

I noticed that your read cache (rcache) is often zero and your write cache (wcache) flops around between zero and 52 averaging 26. What are you running?

[Tom Sekar]

# kmtune -l -q dbc_max_pct
Parameter: dbc_max_pct
Current: 8
Planned: 8
Default: 50
Minimum: -
Module: -
Version: -
Dynamic: No
# kmtune -l -q dbc_min_pct
Parameter: dbc_min_pct
Current: 4
Planned: 4
Default: 5
Minimum: -
Module: -
Version: -
Dynamic: No
# vmstat 2 20
procs memory page faults cpu
r b w avm free re at pi po fr de sr in sy cs us sy id
4 9 0 4646579 279720 50 13 1 0 0 0 5 11128 62268 6340 31 5 64
4 9 0 4646579 279598 29 6 0 0 0 0 0 5285 38422 2795 33 3 64
4 9 0 4646579 279598 19 3 0 0 0 0 0 5298 36721 2761 37 3 59
4 9 0 4646579 279598 12 1 0 0 0 0 0 5184 33397 2690 33 3 64
3 5 0 1991146 279598 71 16 0 0 0 0 0 4995 31936 2587 33 5 62
3 5 0 1991146 279570 46 9 0 0 0 0 0 4934 38537 3202 41 4 55
3 5 0 1991146 279566 30 5 0 0 0 0 0 5014 35950 3090 33 3 64
3 5 0 1991146 279566 19 3 0 0 0 0 0 4939 32680 2860 33 2 65
3 5 0 1991146 279566 12 1 0 0 0 0 0 4956 29711 2674 35 3 62
5 4 0 1987740 279566 66 15 0 0 0 0 0 4917 30645 2827 35 4 62
5 4 0 1987740 279566 42 9 0 0 0 0 0 4892 31223 2977 33 3 64
5 4 0 1987740 279566 27 5 0 0 0 0 0 4962 33578 3045 45 5 50
5 4 0 1987740 279566 14 2 0 0 0 0 0 4885 34835 3245 50 3 47
5 4 0 1987740 279565 9 0 0 0 0 0 0 4723 37231 3470 41 4 55
4 5 0 1977089 279565 52 12 0 0 0 0 0 4658 39147 3618 39 4 57
4 5 0 1977089 278982 34 7 0 0 0 0 0 4645 39397 3690 39 4 56
4 5 0 1977089 278978 21 4 1 0 0 0 0 4677 43378 3915 41 4 54
4 5 0 1977089 278761 14 2 0 0 0 0 0 4735 52519 4449 54 4 42
4 5 0 1977089 278761 9 0 0 0 0 0 0 4800 46314 4092 37 4 59
3 5 0 2168942 278694 166 34 1 0 0 0 0 4609 49554 4364 42 6 52
# swapinfo -atm
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4096 2935 1161 72% 0 - 1 /dev/vg00/lvol2
dev 8192 3048 5144 37% 0 - 1 /dev/vg00/lvswap1
dev 8192 0 8192 0% 0 - 0 /dev/vg00/lvswap2
reserve - 6416 -6416
memory 19248 12000 7248 62%
total 39728 24399 15329 61% - 0 -
#
This box runs SAP and oracle. Thanks much for the response. Looking for your valuable advise.

Thanks

Tom

[Michael Steele_2]

What are nbuf and bufpages?

kmtune -q -l nbuf
kmtune -q -l bufpages

All of your swap priorities should be the same. '1' is the default. Here are the default values for dbc_max_pct 50 and dbc_min_pct 5 but most admins use 7 and 5.

[Tom Sekar]

Both are 0. we are using dynamic

Thanks

Tom

[Darrel Louis]

Tom,

According to the vmstat output the server isn't paging in or out.

Can you please check the following kernel parameters:
Parameters for Memory Paging and Variable Page Size:
maxswapchunks Max swap space on the system 16384
swapmem_on Enable/disable pseudo-swap 1 (on)
swchunk Swap chunk size 4096 (up to 65536 for large RAM)
vps_ceiling Max page size in Kbytes 64

But it looks like you will need to buy some extra memory.

Darrel

[Chauhan Amit]

Hi Tom,

I had a look at vmstat and swapinfo output , here are my observations:
a) There is no Memory pressure on the system
b) memory is 61% utilised which is very much within the limits
c) There is no page-in and page-out happenning , that means there is no pressure on swap as well

If you are facing and bottlenecks then pls mention. Seeing the above output there is no issue which i could find with the system.

-Amit

[Ninad_1]

I dont think you need to add more memory - you already have around 1 GB free memory.
Your main issue is very low buffer hit ratio.
This can be solved depending on what is causing the reads.
If its Oracle/SAP - then try increasing the SGA for Oracle and whatever similar for SAP.
If its some general applications reading a lot of files/data - then try increasing the dbc_max_pct


Regards,
Ninad