General
cancel
Showing results for 
Search instead for 
Did you mean: 

Interpret sar -d sar -u readings, disk % busy

 
SOLVED
Go to solution
Highlighted
Frequent Advisor

Interpret sar -d sar -u readings, disk % busy

(Running on HP-UX 10.20)
I get some high % busy on the disk from time to time, with corresponding low % CPU idle. This usually lasts only for a few seconds. Does this mean that my system is very close to trouble, or is this normal?


sar -d
15:23:48
15:23:49
15:23:50
15:23:51 c2t6d0 25.00 0.50 94 1504 4.50 2.76
15:23:52 c2t6d0 98.00 158.35 555 8880 173.49 8.15
15:23:53 c2t6d0 94.00 0.50 334 5318 5.37 2.74
15:23:54 c2t6d0 97.00 0.50 380 6080 5.00 2.43
15:23:55 c2t6d0 91.00 0.50 363 5808 4.81 2.55
15:23:56 c2t6d0 13.00 0.50 51 816 5.16 2.74
15:23:57
15:23:58 c2t6d0 53.00 200.23 437 6992 214.96 9.62

sar -u
xxxxxxxx %usr %sys %wio %idle
15:23:48 0 0 0 100
15:23:49 0 0 0 100
15:23:50 0 0 0 100
15:23:51 9 3 31 57
15:23:52 8 8 84 0
15:23:53 5 5 90 0
15:23:54 1 2 97 0
15:23:55 7 2 91 0
15:23:56 1 1 4 94
15:23:57 0 0 0 100
15:23:58 0 0 0 100


8 REPLIES 8
Highlighted
Acclaimed Contributor

Re: Interpret sar -d sar -u readings, disk % busy

Hi:

You don't offer enough information to other than comment about possibilities.

I presume that this disk is your boot disk (vg00).

Are you swapping? Look at:

# swapinfo -tam
# vmstat

For 'vmstat' look at the page-out activity. Low numbers are good. Values > 10 indicate memory pressure.

Is your dynamic buffer cache 'dbc_max_pct' too high such that too many file buffers are in memory and when 'syncer' runs (every 30-seconds) your CPU is taxed?

If you have 'glance' that would offer an easy insight.

Regards!

...JRF...
Highlighted
Solution

Re: Interpret sar -d sar -u readings, disk % busy

% busy values tell you very little, in this case they are simply telling you that for several seconds the disks are busty doing something for most of the interval.

What you should pay attention to is avserv and avque numbers - avserv of over 10ms might indicate a problem, and of over 20ms is certainly a problem. You have a fairly heft avque number for 1 second as well, which in turn leads to a fairly large avwait for 1 second as well.

Whether this spells trouble for your server depends what it is doing - what application is running on the system and writing to this disk?

HTH

Duncan

I am an HPE Employee
Accept or Kudo
Highlighted
Frequent Advisor

Re: Interpret sar -d sar -u readings, disk % busy

Thanks for the answer

Yes, this is the boot disk (only 1 disk in system)
Unfortunately I do not have the Glance Sw, it says "license expired"

The vmstat readings and swapinfo looks good as far as I can see. (Page pi and po usually 0)

I have not looked at the dynamic buffer cache, where is this setting?


swapinfo
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 524288 0 524288 0% 0 - 1 /dev/vg00/lvol2
reserve - 204376 -204376

vmstat:

procs memory page faults cpu
r b w avm free re at pi po fr de sr in sy cs us sy id

1 1 0 11206 43173 0 0 0 0 0 0 0 163 1191 67 0 1 99
1 1 0 11206 43173 0 0 0 0 0 0 0 170 2382 72 11 2 87
1 1 0 11206 43173 0 0 0 0 0 0 0 169 1983 65 1 1 98
1 1 0 11206 43173 0 0 0 0 0 0 0 182 2680 76 1 2 97
1 1 0 11206 43173 0 0 0 0 0 0 0 261 2739 186 5 5 90
2 1 0 11742 43173 0 0 0 0 0 0 0 354 3183 317 2 5 93
2 1 0 11742 43173 0 0 0 0 0 0 0 418 3551 420 4 7 89
2 1 0 11742 43173 0 0 0 0 0 0 0 466 3617 494 1 5 94

regards,
D
Highlighted
Honored Contributor

Re: Interpret sar -d sar -u readings, disk % busy

> I have not looked at the dynamic buffer cache, where is this setting?

these are kernel parameters. Please run this command to see what they are set to

kmtune | grep ^dbc

________________________________
UNIX because I majored in cryptology...
Highlighted
Frequent Advisor

Re: Interpret sar -d sar -u readings, disk % busy

Hi,
Could not find the kmtune on my system, is this not included in HP-UX?


The application that is running is a logging application. It is supposed to spread its disk writing out over time, but from these results it seems not to work well.


(A more readable swapinfo:
# swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 512 0 512 0% 0 - 1 /dev/vg00/lvol2
reserve - 199 -199
total 512 199 313 39% - 0 -
Highlighted
Frequent Advisor

Re: Interpret sar -d sar -u readings, disk % busy

I should probably use sysdef instead, on HP-UX 10.20
Highlighted
Frequent Advisor

Re: Interpret sar -d sar -u readings, disk % busy

The Dynamic Buffer Cache is 10%.
Regards,
D

# sysdef
NAME VALUE BOOT MIN-MAX UNITS FLAGS
acctresume 4 - -100-100 -
acctsuspend 2 - -100-100 -
allocate_fs_swapmap 0 - - -
bufpages 13107 - 0- Pages -
create_fastlinks 1 - - -
dbc_max_pct 10 - - -
dbc_min_pct 10 - - -
default_disk_ir 1 - - -
dskless_node 0 - 0-1 -
eisa_io_estimate 1536 - - -
eqmemsize 17 - - -
file_pad 10 - 0- -
fs_async 0 - 0-1 -
hpux_aes_override 0 - - -
maxdsiz 503866 - 256-655360 Pages -
maxfiles 240 - 30-2048 -
maxfiles_lim 2048 - 30-2048 -
maxssiz 20396 - 256-655360 Pages -
maxswapchunks 1024 - 1-16384 -
maxtsiz 262144 - 256-655360 Pages -
maxuprc 150 - 3- -
maxvgs 10 - - -
msgmap 65470464 - 3- -
nbuf 7584 - 0- -
ncallout 1560 - 6- -
ncdnode 150 - - -
ndilbuffers 30 - 1- -
netisr_priority -1 - -1-127 -
netmemmax 43630592 - - -
nfile 17622 - 14- -
nflocks 200 - 2- -
ninode 5728 - 14- -
no_lvm_disks 0 - - -
nproc 1544 - 10- -
npty 160 - 1- -
nstrpty 200 - - -
nswapdev 10 - 1-25 -
nswapfs 10 - 1-25 -
public_shlibs 1 - - -
remote_nfs_swap 0 - - -
rtsched_numpri 32 - - -
sema 0 - 0-1 -
semmap 98238464 - 4- -
shmem 0 - 0-1 -
shmmni 200 - 3-1024 -
streampipes 0 - 0- -
swapmem_on 0 - - -
swchunk 2048 - 2048-16384 kBytes -
timeslice 10 - -1-2147483648 Ticks -
unlockable_mem 6547 - 0- Pages -
Highlighted
Frequent Advisor

Re: Interpret sar -d sar -u readings, disk % busy

Closed