|
|
Async IO bug
Hi,
After the installation of patch set 10.2.0.5, we seem to be hitting
bug 9949948: "Linux: Process spin under ksfdrwat0 if OS Async IO not
configured high enough". And even after several hits it won't die.
According to MOS, the workarounds are:
1) increase aio-max-nr on the OS level
2) disable async_io on the database level
We are on Red Hat 4 and apparently it's not possible to increase aio-
max-nr on this operating system? It's now set to 65536. I'm not really
in favor of disabling async_io for performance reasons...
Anyone else having the same problem and how did you solve this?
Thanks,
Matthias Hoys
|
|
0
|
|
|
|
Reply
|
mhoys
|
9/2/2010 8:19:14 AM |
|
On Thu, 02 Sep 2010 01:19:14 -0700, mhoys wrote:
> We are on Red Hat 4 and apparently it's not possible to increase aio-
> max-nr on this operating system? It's now set to 65536.
Why wouldn't it be possible to increase aio-max-nr?
$>cd /proc/sys/fs
ADPRD1@oracle14 oracle Last rc=0 /proc/sys/fs
$>cat aio-max-nr
300000
ADPRD1@oracle14 oracle Last rc=0 /proc/sys/fs
$>cat /etc/redhat-release
Red Hat Enterprise Linux AS release 4 (Nahant Update 2)
ADPRD1@oracle14 oracle Last rc=0 /proc/sys/fs
--
http://mgogala.byethost5.com
|
|
0
|
|
|
|
Reply
|
gogala.mladen (1073)
|
9/2/2010 12:40:30 PM
|
|
On Sep 2, 8:40=A0am, Mladen Gogala <gogala.mla...@gmail.com> wrote:
> On Thu, 02 Sep 2010 01:19:14 -0700, mhoys wrote:
> > We are on Red Hat 4 and apparently it's not possible to increase aio-
> > max-nr on this operating system? It's now set to 65536.
>
> Why wouldn't it be possible to increase aio-max-nr?
>
> $>cd /proc/sys/fs
> =A0ADPRD1@oracle14 oracle Last rc=3D0 /proc/sys/fs
> =A0$>cat aio-max-nr
> 300000
> =A0ADPRD1@oracle14 oracle Last rc=3D0 /proc/sys/fs
> =A0$>cat /etc/redhat-release
> Red Hat Enterprise Linux AS release 4 (Nahant Update 2)
> =A0ADPRD1@oracle14 oracle Last rc=3D0 /proc/sys/fs
>
> --http://mgogala.byethost5.com
Lots of places run RedHat and have async IO in use. I would double
check all the kernal parameters against the Installation Guide
requirements.
HTH -- Mark D Powell --
|
|
0
|
|
|
|
Reply
|
Mark.Powell2 (174)
|
9/2/2010 2:26:24 PM
|
|
On Sep 2, 2:40=A0pm, Mladen Gogala <gogala.mla...@gmail.com> wrote:
> On Thu, 02 Sep 2010 01:19:14 -0700, mhoys wrote:
> > We are on Red Hat 4 and apparently it's not possible to increase aio-
> > max-nr on this operating system? It's now set to 65536.
>
> Why wouldn't it be possible to increase aio-max-nr?
>
> $>cd /proc/sys/fs
> =A0ADPRD1@oracle14 oracle Last rc=3D0 /proc/sys/fs
> =A0$>cat aio-max-nr
> 300000
> =A0ADPRD1@oracle14 oracle Last rc=3D0 /proc/sys/fs
> =A0$>cat /etc/redhat-release
> Red Hat Enterprise Linux AS release 4 (Nahant Update 2)
> =A0ADPRD1@oracle14 oracle Last rc=3D0 /proc/sys/fs
>
> --http://mgogala.byethost5.com
You are right! It can be changed on Red Hat 4.
Guess I have been confusing AIO-MAX-NR with AIO-MAX-SIZE! Stupid me...
Matthias
|
|
0
|
|
|
|
Reply
|
matthias.hoys (92)
|
9/2/2010 2:35:55 PM
|
|
|
3 Replies
413 Views
(page loaded in 0.147 seconds)
Similiar Articles: Is this a bug? - comp.soft-sys.matlabAsync IO bug - comp.databases.oracle.server Hi, After the installation of patch set 10.2.0.5, we seem to be hitting bug 9949948: "Linux: Process spin under ksfdrwat0 if OS ... Oracle SQL bug in 9.2.0.8 - comp.databases.oracle.server ...Async IO bug - comp.databases.oracle.server Oracle SQL bug in 9.2.0.8 - comp.databases.oracle.server ... Looks like Oracle f*cked up and re-released patch set 10.2.0.5 for ... Average wait time per IO request and ... - comp.unix.solaris ...... comp.os.linux.misc High I/O Wait times just mean you have spare ... The man page says: await: The average time (in milliseconds) for I/O requests ... Async IO bug - comp ... rsubmit in macro - comp.soft-sys.sasAsync IO bug - comp.databases.oracle.server Is this a macro bug??? - comp.soft-sys.sas Async IO bug - comp.databases.oracle ... the installation of patch set 10.2.0.5, we ... ntpd on AIX6.1 does not start - comp.protocols.time.ntpAsync IO bug - comp.databases.oracle.server ntpd on AIX6.1 does not start - comp.protocols.time.ntp... sockets(123) 18 Jan 17:41:35 ntpd[6226166]: ntp_io ... dev/refclock ... 10.2.0.5 Patchset - comp.databases.oracle.serverLooks like Oracle f*cked up and re-released patch set 10.2.0.5 for ... sys.sas Async IO bug - comp.databases.oracle.server Hi, After the installation of patch set 10.2.0.5 ... file copy: invalid argument - comp.lang.tcl... and aio_write - comp.unix.programmer hello all, I'm trying to copy a file using async io ... Bug#605728: smb/gvfs-copy: Error while copying file, "Invalid ... Package ... How to bound a Cores generated output in Modelsim - comp.arch.fpga ...alterauser wrote: > Yuu must compile it, just like any other hdl file. > > Assuming you have a simulation wrapper for your design containing all > IO parts of the FPGA ... High await in iostat? - comp.os.linux.miscOn the other hand asynchronous i/o would run just fine. > I took a look at collectl ... Average wait time per IO request and ... - comp.unix.solaris ... High await in ... NMEA ref.clock better than my ISP's timeserver? - comp.protocols ...... at redhat dot com This is beta software, so report to him if you find bugs. ... Com Nmea Port Discussions - Forum Search - Omgili Multiple port asynchronous IO over ... Bug 611205 – Use async IO properly, depending on the storage typeRed Hat Bugzilla – Bug 611205. Use async IO properly, depending on the storage type. Last modified: 2011-05-19 09:45:36 EDT Async IO using with FileStream is not async | Microsoft ConnectUser(s) can reproduce this bug ... BeginXXX methods reveals that they do not satistfy most of the requirements of async IO ... 7/24/2012 5:10:25 AM
|
|
|
|
|
|
|
|
|