| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Mailing Lists -> Oracle-L -> hp-ux, oracle 8.x and async io
listers,
there have been numerous postings along the line of does/doesn't work
with the following components
hp-ux 10,11
oracle 8.x
async io
and the fact that it used to under oracle 7.3.4.x
we migrated some months ago from 7.3.4.3.1 to 8.0.6 and took a 20% hit in performance.
oracle have been doing the old 'send us a test case and we'll get the developers to review it' scenario, but don't appear to be being very pro-active about it especially the amount of heat the topic gets on technet etc
basically if you r running the above combination, you will find that no matter how many types you relink the kernel, create the device file that the /dev/async devices timestamp only appears to be updated on a shutdown of the instance.
i've traced the calls and cudn't see anything that referred to aio/async calls during normal operation.
found in an old sybase install script (1995)that there is a problem with the device being created with a minor of 0 and that it should be modified.
if you are having any of the above symptoms and can do the following, id' be please to hear your results.
rm /dev/async
/etc/mknod /dev/async c 101 7
chown oracle:dba /dev/async
chmod 660 /dev/async
with this new value for the minor number, you should see the timestamp being modified as expected, that is every time its used as was the oracle 7 behaviour.
any findings would be most appreciated.
kc
 |
 |