Unable to remove ASM disks from ASMLIB

Hi,
Environment : ASM 11.1.0.7 on RHEL5. I am getting following error while
trying to release the disks from ASMLIB. The disks have been already dropped
from ASM diskgroup.
[root]# /etc/init.d/oracleasm deletedisk VOL0
Removing ASM disk "VOL0": [FAILED]
"Clearing disk header: oracleasm-write-label: Unable to open device
"/dev/oracleasm/disks/VOL0": Device or resource busy failed Unable to clear
disk "VOL0"
It seems that I may be hitting ASM bug 7225720 which states that some Oracle
processes may still have these devices opened. I could resolve the issue by
stopping the ASM as support has suggested, but I am hoping to find a
solution with taking outage on production server.
Martin Meyer has mentioned a workaround on his blog
http://martinmeyer.blogspot.com/2008/12/remove-asm-disk.html by creating and
dropping a junk diskgroup on the dropped disks, but it didn't work for me. I
am wondering if anyone has experienced this issue and found any other
workaround?
Thanks
Manmohan

Use this to find who is holding it open. (Also it could be locked on other
nodes to, so check all nodes).
lsof | grep /dev/oracleasm
On Tue, Jan 26, 2010 at 4:54 PM, Manmohan Jalsingh wrote:
> Hi,
>
> Environment : ASM 11.1.0.7 on RHEL5. I am getting following error while
> trying to release the disks from ASMLIB. The disks have been already dropped
> from ASM diskgroup.
>
> [root]# /etc/init.d/oracleasm deletedisk VOL0
> Removing ASM disk "VOL0": [FAILED]
>
> "Clearing disk header: oracleasm-write-label: Unable to open device
> "/dev/oracleasm/disks/VOL0": Device or resource busy failed Unable to clear
> disk "VOL0"
>
> It seems that I may be hitting ASM bug 7225720 which states that some
> Oracle processes may still have these devices opened. I could resolve the
> issue by stopping the ASM as support has suggested, but I am hoping to find
> a solution with taking outage on production server.
>
> Martin Meyer has mentioned a workaround on his blog
> http://martinmeyer.blogspot.com/2008/12/remove-asm-disk.html by creating
> and dropping a junk diskgroup on the dropped disks, but it didn't work for
> me. I am wondering if anyone has experienced this issue and found any other
> workaround?
>
> Thanks
> Manmohan
>
>
>
--
Thomas Roach
813-404-6066
troach@gmail.com

We had a similar issue with oracle 10g not closing the file handles on
dropped disks. In our case we are using raw devices.
After making sure that the disks were indeed dropped we decided to remove
the devices, ignoring the open files.
It didn't have any bad consequences. We have done so probably more than ten
times and it always worked.
Harel
=D7=91-26/01/2010, =D7=91=D7=A9=D7=A2=D7=94 23:54, Manmohan Jalsingh =D7=9B=D7=AA=D7=91/=D7=94:
Hi,
Environment : ASM 11.1.0.7 on RHEL5. I am getting following error while
trying to release the disks from ASMLIB. The disks have been already droppe=
d
from ASM diskgroup.
[root]# /etc/init.d/oracleasm deletedisk VOL0
Removing ASM disk "VOL0": [FAILED]
"Clearing disk header: oracleasm-write-label: Unable to open device
"/dev/oracleasm/disks/VOL0": Device or resource busy failed Unable to clear
disk "VOL0"
It seems that I may be hitting ASM bug 7225720 which states that some Oracl=
e
processes may still have these devices opened. I could resolve the issue b=
y
stopping the ASM as support has suggested, but I am hoping to find a
solution with taking outage on production server.
Martin Meyer has mentioned a workaround on his blog
http://martinmeyer.blogspot.com/2008/12/remove-asm-disk.html by creating an=
d
dropping a junk diskgroup on the dropped disks, but it didn't work for me. =
I
am wondering if anyone has experienced this issue and found any other
workaround?
Thanks
Manmohan

Thomas,
It is a non-rac system. I ran "lsof |grep /dev/oracleasm" which returned a
long list. Is there a way
to identify which processes are connected to dropped devices ?
# lsof | grep /dev/oracleasm
oracle 367 oracle 9u REG 0,25 0 18446604573976970728 /dev/oracleasm/iid/
0000000000000006
oracle 383 oracle 9u REG 0,25 0 18446604573864723880
/dev/oracleasm/iid/0000000000000010
oracle 451 oracle 11u REG 0,25 0 18446604573864723880
/dev/oracleasm/iid/0000000000000010
oracle 517 oracle 9u REG 0,25 0 18446604497873475624
/dev/oracleasm/iid/000000000000000C
oracle 835 oracle 9u REG 0,25 0 18446604497873474344
/dev/oracleasm/iid/000000000000000D
oracle 838 oracle 9u REG 0,25 0 18446604497873474344
/dev/oracleasm/iid/000000000000000D
oracle 840 oracle 9u REG 0,25 0 18446604497873474344
/dev/oracleasm/iid/000000000000000D
oracle 842 oracle 9u REG 0,25 0 18446604497873474344
/dev/oracleasm/iid/000000000000000D

Thanks
Manmohan
On Tue, Jan 26, 2010 at 2:18 PM, Thomas Roach wrote:
> Use this to find who is holding it open. (Also it could be locked on other
> nodes to, so check all nodes).
>
> lsof | grep /dev/oracleasm
>
>
> On Tue, Jan 26, 2010 at 4:54 PM, Manmohan Jalsingh wrote:
>
>> Hi,
>>
>> Environment : ASM 11.1.0.7 on RHEL5. I am getting following error while
>> trying to release the disks from ASMLIB. The disks have been already dropped
>> from ASM diskgroup.
>>
>> [root]# /etc/init.d/oracleasm deletedisk VOL0
>> Removing ASM disk "VOL0": [FAILED]
>>
>> "Clearing disk header: oracleasm-write-label: Unable to open device
>> "/dev/oracleasm/disks/VOL0": Device or resource busy failed Unable to clear
>> disk "VOL0"
>>
>> It seems that I may be hitting ASM bug 7225720 which states that some
>> Oracle processes may still have these devices opened. I could resolve the
>> issue by stopping the ASM as support has suggested, but I am hoping to find
>> a solution with taking outage on production server.
>>
>> Martin Meyer has mentioned a workaround on his blog
>> http://martinmeyer.blogspot.com/2008/12/remove-asm-disk.html by creating
>> and dropping a junk diskgroup on the dropped disks, but it didn't work for
>> me. I am wondering if anyone has experienced this issue and found any other
>> workaround?
>>
>> Thanks
>> Manmohan
>>
>>
>>
>
>
> --
> Thomas Roach
> 813-404-6066
> troach@gmail.com
>