Hi Hertha/Gerrard/anyone, SAN disk partitions device files changes with each reboot

Tue May 27 09:11:37 UTC 2008

Hi Gerrard,

So ultimately which one should we use?
/dev/mapper/...     or   ...

My apologies, I'm a bit confused after reading
the reply

 Guess I have to read up more about this at
http://www.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/5.2/html/DM_Multipath/index.html
which you gave me previously but haven't got a chance.

Thanks
Goh

On 5/27/08, Gerrard Geldenhuis <Gerrard.Geldenhuis at datacash.com> wrote:
>
> Hi,
> I asked a similar question a while ago and I copy in the response. I hope
> it helps. Thanks to Christophe Varoqui for supplying the answers.
>
> Le lundi 04 février 2008 à 12:50 +0000, Gerrard Geldenhuis a écrit :
> > Hi Christophe,
> >
> > I am a bit confused between the usage of
> >
> > /dev/mpath
> >
> > /dev/dm-X and
> >
> > /dev/mapper/
> >
> >
> >
> > I am unsure as to which device I should be using when creating lvm
> > volumes. I have asked a consultant from Redhat who gave the following
> > response:
> >
> 
> You can use pvcreate on whatever object you want. The important setup is
> the "filter" in /etc/lvm/lvm.conf : this filter has to use only one kind of
> object to avoid pvscan confusion. Here again, whatever naming convention is
> good.
>
> The /dev/mpath/ contains only multipath-type devmaps, whereas /dev/mapper
> contains all kind of devmap types (linear for LV)
>
> >
> > We ended up having a big discussion about this on IRC yesterday, and
> > the outcome was inconclusive. However, the guy who is our oracle/san
> > expert says /dev/mpath, so that is what I'd do.
> >
> Also keep in mind that you can disable user_friendly_names in
> multipath.conf, which will give you /dev/{mapper,mpath}/6000111122223333
> names.
> Those are really interesting when you use clusters (like RAC) because the
> naming is consistent between hosts.
>
> >
> >
> > You might also find it useful to take a look at the kpartx command,
> > and use that after you've added a partition to a LUN. It should see to
> > it that the relevant /dev/mpath partition device gets created without
> > having to reboot the system
> >
> Avoid partitioning multipathed device when possible : it will remove
> considerable complexity to the software stack.
> >
> >
> > I also asked on rhel5 mailinglist where I got the following response:
> >
> >
> >
> > Using /dev/mapper is always how I've seen it done. /dev/mpath/* looks
> > to be just a symlink to /dev/dm-? device nodes which are, in turn,
> > device nodes with identical major/minor numbers as /dev/mapper/*.
> >
> >
> >
> > Why /dev/mpath/* is even there, I'm not sure.
> >
> >
> udev rules trigger their creation
>
> Regards,
> cvaroqui
>
> > -----Original Message-----
> > From: redhat-list-bounces at redhat.com [mailto:redhat-list-
> > bounces at redhat.com] On Behalf Of sunhux G
> > Sent: 26 May 2008 10:37
> > To: General Red Hat Linux discussion list
> > Subject: Hi Hertha/Gerrard/anyone,SAN disk partitions device files
> changes
> > with each reboot
> >
> > Hi Hertha/Gerrard/Anyone else,
> >
> >
> > Thanks for the previous excellent replies to my questions.
> >
> > Something new just surfaced with the NetApp SAN disks partitions
> > that are presented to our RHES 4.6 :
> >
> > The current mappings on 1st server is :
> > lrwxrwxrwx  1 root root 7 May 22 15:36 mpath0 -> ../dm-2
> > lrwxrwxrwx  1 root root 7 May 22 15:36 mpath1 -> ../dm-5
> > lrwxrwxrwx  1 root root 7 May 22 15:36 mpath2 -> ../dm-3
> > lrwxrwxrwx  1 root root 7 May 22 15:36 mpath3 -> ../dm-4
> > lrwxrwxrwx  1 root root 7 May 22 15:36 mpath4 -> ../dm-1
> > lrwxrwxrwx  1 root root 7 May 22 15:36 mpath5 -> ../dm-0
> >
> > & "multipath -ll" gives :
> >
> > mpath0 (360a98000567244396334493370345055)
> > [size=5 GB][features="1 queue_if_no_path"][hwhandler="0"]
> > \_ round-robin 0 [active]
> >  \_ 8:0:2:5 sds 65:32  [active]
> >  \_ 8:0:3:5 sdy 65:128 [active]
> > \_ round-robin 0 [enabled]
> >  \_ 8:0:1:5 sdm 8:192  [active]
> >  \_ 8:0:0:5 sdg 8:96   [active]
> >
> > mpath1 (360a9800056724439633449336c786d69)
> > [size=5 GB][features="1 queue_if_no_path"][hwhandler="0"]
> > \_ round-robin 0 [active]
> >  \_ 8:0:2:4 sdr 65:16  [active]
> >  \_ 8:0:3:4 sdx 65:112 [active]
> > \_ round-robin 0 [enabled]
> >  \_ 8:0:0:4 sdf 8:80   [active]
> >  \_ 8:0:1:4 sdl 8:176  [active]
> >
> > On another  Linux server (with cluster script /etc/init.d/o2cb_start.sh
> > started),
> > /dev/mpath/mpath0 or mpath1 or ... mpathx completely maps to different
> > minor devices  /dev/sd...
> >
> > So we mounted on server 1 a partition (formatted as ocfs using
> > ocfs2console)
> > first & create a test file on it & then on server 2, we mount
> > mpath0/.../mpathx
> > one after another to see which of it has the test file on it to identify
> > it.
> >
> > We then put in vfstab the /dev/mpath/mpathx & mountpoint for each server
> > that we have determined the hard way.
> >
> > Alas, after we rebooted both the servers, all the mappings became
> > different
> > ie on server 1 where mpath0 ->../dm-2 became mpath0 ->../dm-4 after
> reboot
> > & on server 2 where mpath0 ->../dm-1 became mpath0 ->../dm-3 after
> reboot.
> >
> > Oracle told us to use /dev/mapper/mpathx  - this appears to be more
> > reliable
> > (ie it's fixed to a specific partition regardless of how many reboots are
> > done).
> >
> > Can someone explain what's the differences between
> >  /dev/mpath/mpathx   &  /dev/dm-x   &   /dev/mapper/mpathx
> >
> > Or I've not completely installed all the required stuff from NetApp on
> our
> > Redhat servers yet that triggers this?
> >
> >
> > Thanks
> > U
> > --
> > redhat-list mailing list
> > unsubscribe mailto:redhat-list-request at redhat.com?subject=unsubscribe
> > https://www.redhat.com/mailman/listinfo/redhat-list
>
> --
> redhat-list mailing list
> unsubscribe mailto:redhat-list-request at redhat.com?subject=unsubscribe
> https://www.redhat.com/mailman/listinfo/redhat-list
>