Re: Oracle on raw devices in SLES9

Shashi Kanth Boddula <[email protected]>
Newsgroups gmane.linux.suse.oracle.general,gmane.linux.suse.sles-e
Message-ID <[email protected]>
Michael, Thanks for your reply.

Your given solution worked for me. Thank you once again.

I always thought that i need to touch multipath only when if i have 
multiple paths to a LUN. But in my case , i don't have multiple paths, 
so i thought multipath is not the way to go. But, today i learned that 
it can also used for persistent binding.

But, still i have one more problem with SAN attached disks.

I am creating oracle database on raw in two ways

(1) LVM on a LUN
(2) Physical partitions on a LUN.

i am successful at (2) but not (1)

In (2), i have partitioned the LUN into several partitions , and added 
that LUN into multipath.conf, and added all those /dev/mapper/*  
partitions to my /etc/raw , and installed oracle database, and it 
worked, it installed successfully.

Now , In (1) , Created several LVs on a LUN, added that LUN to 
multupath.conf, configured lvm.conf to bring LVM under the control of 
multipath, and added all /dev/mapper/LV* in /etc/raw , and tried to 
installed oracle database, but it failed. In the "dbca", "step 10 of 
11", "storage section" , it throws an error saying that "Raw device 
validation check for data file /dev/raw/raw1" failed, file /dev/raw/raw1 
is not a raw device . "Device or resource busy" .

Here, anyone, please tell me, RAW-Linux LVM-Multipath  is actually 
supported?

Any help here ?

But, i can tell one thing, in (1) , without multipath , and by giving 
all /dev/volume_group/LV* to /etc/raw , it works , and  it worked for me.

-- Shashi



Michael Kershaw wrote:
> First off, I would never listen to anything REDHAT has to say about
> anything...seriously...maybe just my personal opinion, but it comes
> from a lot of experience between the two big commercial players out
> there...ok..with that said, I have multiple Sles9 Sp3, 10g clusters
> running at the moment.  Some 2 node clusters,  all the way up to
> multiple 18 node clusters in production.  They all utilize shared
> fiber storage, and raw devices with ASM.  I use the device-mapper for
> my persistent binding, and udev for the persistent permissions.  I've
> created a multipath.conf file aliasing my devices to the UID of each
> of the multipath'd luns.  Here's an entry for an individual lun:
>
> multipath {
>                wwid                    1HITACHI_R450A82C09A5
>                alias                     ora_registry1
>                path_grouping_policy    multibus
>                path_checker            readsector0
>                path_selector           "round-robin 0"
>                failback                60
> }
>
> and then make entries in my /etc/raw file like so:
>
> ##
> # First 5 devices for Oracle services
> ##
> raw1:mapper/ora_voting1p1
> raw2:mapper/ora_voting2p1
> raw3:mapper/ora_voting3p1
> raw4:mapper/ora_registry1p1
> raw5:mapper/ora_registry2p1
>
>
> The p1 obviously is the partition for the device of the alias created.
> I simply lay out a partition on each of the luns before I pull them
> into the device-mapper for consistency, and that way anyone else in my
> team can easily see what devices are free on the system in question if
> there is no partition on said device when they're looking.  From the
> /etc/raw file, I just have my entries for permissions in
> /etc/udev/udev.permissions and away we go.  I've had our larger
> clusters running in this fashion for well over a year now.  If you
> have any more questions on anything more specific, or If I've confused
> you even more, just let me know and I can get into more detail.
>
>
> Hope this helps,
> Mike
>
>
>
>
>
> On 6/2/07, Shashi Kanth Boddula <[email protected]> wrote:
>> Ok , the problem "persistence ownership" on raw volumes is over by using
>> udev, now i am facing a one more problem in SLES9 SP3, and the problem
>> is with SAN disks (raw disks) (persistent binding).
>>
>> I am trying to solve the problem by using same udev concept, as i found
>> that we need add line in the udev rules file, and redhat documentation
>> said that we need add a line in /etc/udev/rules.d/20-names.rules file.
>>
>> http://kbase.redhat.com/faq/FAQ_85_8082.shtm
>>
>> (KERNEL="sd*", BUS="scsi", PROGRAM="/sbin/scsi_id", RESULT="UUID",
>> NAME="devicenam
>>
>> As SLES9 ships a single udev rules file, i have added the above line in
>> /etc/udev/udev.rules in SLES9.
>>
>>
>> KERNEL="sd*", BUS="scsi", PROGRAM="/sbin/scsi_id",
>> RESULT="3600508b4001007d000017000044a0000", NAME="disk/fc/fc_lun0"
>>
>> For each LUN, i have added a line like above.
>>
>> Now in the YaST, when i run (#yast2 disk) , in the YaST GUI, i am seeing
>> all my assigned names.
>>
>> /dev/disk/fc/fc_lun0
>>
>> /dev/disk/fc/fc_lun2
>>
>> /dev/disk/fc/fc_lun3
>>
>> so on..................
>>
>>
>> But in any LUN, if i try to create a partition, YaST doesn't through any
>> error, but actually the partition will not be created.
>>
>> when you run YaST next time, that partitions own't be there .
>>
>> Any one has any solution for this?
>>
>> How others are doing ?
>>
>>
>> Thanks
>>
>>
>>
>> Alexei_Roudnev wrote:
>> > If you use standard SuSe orarun, then oracle get a secondary group 
>> 'disk'
>> > and have read/write access to the raw devices by default. It works for
>> > everything except CSSFile on CRS (RAC cluster).
>> >
>> > Other method is, yes, to modify udev permissions.
>> >
>> > But I use modified rawnames script which
>> > - creates symlinks for raw devices so that oracle dont work with 
>> 'raw' names
>> > but works with names such as 'DISK1', 'LOG2' and so on;
>> > - it uses 'sleep 5; chown' so it set up owner after udev finish t's 
>> work.
>> >
>> >
>> >
>> > ----- Original Message -----
>> > From: "Kevin Lyons" <[email protected]>
>> > To: "Shashi Kanth Boddula" <[email protected]>
>> > Cc: <[email protected]>; <[email protected]>
>> > Sent: Friday, June 01, 2007 9:36 AM
>> > Subject: Re: [suse-oracle] Oracle on raw devices in SLES9
>> >
>> >
>> >
>> >> On Fri, 2007-06-01 at 21:37 +0530, Shashi Kanth Boddula wrote:
>> >>
>> >>> Hi All,
>> >>>
>> >>> I am installing Oracle 10gR2 on SLES9 SP3 raw volumes which is 
>> made by
>> >>> Linux LVM. As per the oracle documentation, all the raw devices 
>> should
>> >>> be owned by oracle user, for that we need to change the ownership 
>> on all
>> >>> raw devices  where your installing database to oracle user.
>> >>>
>> >>> The problem is , after you create a mapping file in /etc/raw, and 
>> when
>> >>> you start the raw service (#rcraw start), the permission again 
>> set back
>> >>> to root.
>> >>>
>> >>> I have tried to solve this problem by modifying /etc/init.d/raw 
>> file ,
>> >>> but still , always , when you start the raw services, the 
>> ownership on
>> >>> raw devices again go back to root.
>> >>>
>> >>> Anyone has any solution for this ?
>> >>>
>> >>>
>> >> Modifying the /etc/init.d/raw script is not the way to go.
>> >>
>> >> When the raw devices (or really any devices) are created, the udev
>> >> subsystem automatically sets the permissions to root:disk:660.
>> >>
>> >> On SLES9 edit the /etc/udev/udev.permissions file which has rules for
>> >> what permissions are given to devices files as they are created.
>> >>
>> >> For raw devices, the permissions are set by the following line:
>> >>
>> >> raw/raw*:root:disk:660
>> >>
>> >> You want to add some rules to set the permissions for the raw devices
>> >> that you are going to use like so:
>> >>
>> >> raw/raw1:oracle:oinstall:644
>> >> raw/raw2:oracle:oinstall:644
>> >> raw/raw3:oracle:oinstall:644
>> >> raw/raw*:root:disk:660
>> >>
>> >> Or whatever is appropriate for your setup.  The rules can contain 
>> some
>> >>
>> > standard
>> >
>> >> shell expansions like 'raw/raw[1-3]' or similar.  The man page has 
>> more
>> >>
>> > details.
>> >
>> >> Note that the file is processed in order, and when a matching rule is
>> >>
>> > found
>> >
>> >> processing stops.  So you need to have the default rule come after 
>> your
>> >>
>> > custom
>> >
>> >> rules or it will not work.
>> >>
>> >> Thanks,
>> >> Kevin
>> >>
>> >>
>> >>
>> >> --
>> >> To unsubscribe, email: [email protected]
>> >> For additional commands, email: [email protected]
>> >> Please see http://www.suse.com/oracle/ before posting
>> >>
>> >>
>> >>
>> >
>> >
>>
>>
>> -- 
>> To unsubscribe, email: [email protected]
>> For additional commands, email: [email protected]
>> Please see http://www.suse.com/oracle/ before posting
>>
>>


-- 
To unsubscribe, email: [email protected]
For additional commands, email: [email protected]
Please see http://www.suse.com/oracle/ before posting
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.