Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: Corrupted Datablocks with Oracle under Linux
Hi,
May be we have found a solution for this problem.
Disable the asynchronous mode for disk and tape access in the init.ora file.
(disk_asynch_io=FALSE, ...).
Since these changes, all systems are working stable without any errors - but
this costs performance !
BTW: Informix for Linux does not support the asynchro disk mode
http://www.informix.com/informix/products/linux/lxfaq.html#idsfaq
"Informix Dynamic Server , Linux Edition Suite has a feature set that is specifically geared to the needs of the Linux community, such as the Apache Integration HOWTO, RPM, ESQL/C, ODBC,etc. At this time, asynchronous I/O is not supported in the Linux kernel."
This description was the reason to disable this I/O mode for the oracle engine.
Theo van der Kruis <theo_at_incore.xs4all.nl> schrieb in Nachricht
x2k8oi9fak.fsf_at_incore.xs4all.nl...///
>>>>>> "Thomas" == Thomas Winterschlade <thomas_at_imagedev.com> writes:
>
> Thomas> Hi ! We have serious troubles using Oracle 8.0.5.0 /
> Thomas> 8.0.5.1 under Linux (Kernel 2.2.9, SuSE distribution 6.2).
>
> Thomas> From time to time the database crashes with the Messages
> Thomas> "ORA-01578: ORACLE data block corrupted (file # x, block #
> Thomas> y)".
>
>We have similar problems with Oracle 8.0.5.0 on Linux kernel 2.0.36 (I
>can't check right now, it's the standard kernel that comes with RedHat
>5.2). It happens about once a week on two development machines,
>especially under heavy load. The machines have 128MB RAM each and 2
>swap partitions of 128MB. I've added a 128MB swapfile to each machine
>because I've read somewhere 3 times physical mem was recommended, but
>it didn't help.
>
>In the same document that talked about the swapsize there was also a
>comment to increase the maximum amount of shared memory. I haven't
>done that yet. We've talked to Oracle support here in the Netherlands
>and they're thinking of a Linux bug, which I now tend to agree with
>because we haven't got any problems with a similar machine running NT
>(although it has Oracle 8.0.5.1 instead of 8.0.5.0). Oracle support
>is trying to test it on their own Linux system but are having problems
>installing it.
>
>Theo.
>
Received on Thu Oct 21 1999 - 03:55:09 CDT