Help on vmstat -v / AIX 5.2 1

[dbinsight]

we have I/O issue on the AIX box for our Oracle DB
the disks having the Database files are always 100% busy
and the wa column in vmstat hits above 50

kthr memory page faults cpu
----- ----------- ------------------------ ------------ -----------
r b avm fre re pi po fr sr cy in sy cs us sy id wa
2 7 3794904 92069 0 3 5 58 107 0 1168 19720 10925 19 4 48 29
1 3 3794866 92110 0 0 1 1522 62283 0 1850 15572 3881 9 3 29 60
and the vmstat -v show the I/O's being blocked
2238141 pending disk I/Os blocked with no pbuf
13963233 paging space I/Os blocked with no psbuf
2740 filesystem I/Os blocked with no fsbuf
1423313 client filesystem I/Os blocked with no fsbuf
1128548 external pager filesystem I/Os blocked with no fsbuf

What does these indicate, short of real mem or does some kernal parameters need to be adjusted?

 

[p5wizard]

I'd up the number of fsbufs per filesystem.


What are your current settings?

ioo -L|egrep 'numfsbufs|j2_nBufferPerPagerDevice'

numfsbufs is for jfs filesystems
j2_nBuffer... is for jfs2 filesystems

if I'm not mistaken.

Note, if you change these values, you have to umount/mount the filesystems to take effect. I.e. you have to bring Oracle down.


HTH,

p5wizard

 

[dbinsight]

p5wizard, Thanks, I dont have the access to it,i will get the SA to get me the output.
Are these figures cummulative since the last reboot of the box.
what a good setting for this

 

[dbinsight]

ioo -L |egrep 'numfsbufs|j2_nBufferPerPagerDevice'
numfsbufs 512 196 512 1 2G-1 M
j2_nBufferPerPagerDevice 512 512 512 0 2G-1

The above are our settings, Are these the default settings?

 

[p5wizard]

answers to your questions:

yes, cumulative (so depends on how long the system's been running to interpret the values).

no, already been increased for jfs
yes, defaults for jfs2

1st value = current situation
2nd value = system default
3rd value = value for nextboot

# ioo -L|head -3; ioo -L|egrep 'numfs|j2_nBuff'
NAME CUR DEF BOOT MIN MAX UNIT TYPE
DEPENDENCIES
--------------------------------------------------------------------------------
j2_nBufferPerPagerDevice 512 512 512 0 256K M
numfsbufs 400 196 400 1 2G-1 M

But as I said before, doesn't help to increase 'em unless you unmount/mount the filesystems. As your SA has upped the 'NEXTBOOT' values, I guess (s)he knows about that.

Run "topas 2" for a few iterations, and post that screenful please.
Also "vmo -L|egrep 'perm%|client%'" output please.

You have a very high value:
13963233 paging space I/Os blocked with no psbuf
On my large DB servers this is close to zero.

Run "lsps -a" and post that output also please.

I googled for "aix psbufs" and found an Oracle AIX performance Technical Brief, here's an excerpt:

# vmstat –v | tail -5 (we only need last 5 lines)
0 pending disk I/Os blocked with no pbuf
o for pbufs, increase pv_min_pbuf using ioo,
0 paging space I/Os blocked with no psbuf
o for psbufs, stop paging or add more paging space,
8755 filesystem I/Os blocked with no fsbuf ?? JFS
o for fsbufs, increase numfsbufs using ioo,
o default is 196, recommended starting value is 568,
0 client filesystem I/Os blocked with no fsbuf (NFS/Veritas)
o for client filesystem fsbufs, increase:
• nfso’s nfs_v3_pdts and nfs_v3_vm_bufs
2365 external pager filesystem I/Os blocked with no fsbuf (JFS2)
o for external pager fsbufs, increase:
• j2_nBufferPerPagerDevice, default is 512, recommended value is 2048,
• j2_dynamicBufferPreallocation using ioo.



HTH,

p5wizard