Re: Direct NFS and ZS3
Date: Tue, 05 May 2015 20:54:10 +0200
Message-ID: <1430852050.10178.45.camel_at_dhoogfr-lpt1>
dNFS should consume less cpu then iSCSI and I find it easier to use. But I don't know why you would create ASM on top of the NFS volumes? Unless you have to use ASM because you have a standard edition rac, it seems to me as an unnecessary layer.
With 11.2 you still need to set your record size to match the db block size and research the zfs volume properties. But mind you, support for 11.2.0.4 will end this year (unless you pay an additional maintenance fee).
One thing that you might want to test is what happens when the NFS
volume containing the OCR becomes offline.
I have tested this with Netapp storage and oracle 10.2 and 11.2 (or .1
can't find it directly), and the io would not failover to the OCR mirror
(as the IO to the voting files does).
This resulted in a halt of all cluster operations (although normal db
operations could continue).
regards,
On di, 2015-05-05 at 11:16 -0500, Fernando Jose Andrade wrote:
> Thanks Freek
>
>
>
> As far as I see it’s a 10Gbps ethernet by fiber connector.
> The LUNs are provisioned by iSCSI.
>
>
> The database is 11.2.0.4, but it’s not on production yet.
>
>
> I don’t know if this configuration is worst than DNFS, I have read
> around the web that DNFS beats iSCSI
> but I haven’t tried response times with ORION or SLOB yet.
>
>
> I wonder.. if I use ASM over DNFS I make zero filled files with dd and
> then I feed ASM with them as disks?
> I have read Yuri's Velikanov post about no penalty in using ASM over
> DNFS but the configuration is not as clear as I wanted.
> (http://www.pythian.com/blog/are-there-performance-penalties-running-oracle-asm-on-nfs/)
>
>
> Thanks for your time.
>
>
> FJA
>
>
> > On 05 May 2015, at 02:15, Freek D'Hooge <freek.dhooge_at_gmail.com>
> > wrote:
> >
> >
> >
> > reposting for list
> >
> > On di, 2015-05-05 at 07:32 +0200, freek wrote:
> >
> > > Fernando,
> > >
> > > You can use dNFS directly, without needing ASM on top of it.
> > > For ocr / voting disks you can use "normal" kernel NFS.
> > >
> > > If you are using oracle 12c, you should have a look at the
> > > nfs_version parameter in the oranfstab to set it to "nfsv4.1",
> > > which will allow the appliance to set the record size correctly
> > > for the different types of files.
> > > (But check first if solaris is supported).
> > >
> > > One question though: how much bandwidth to you have now with the
> > > FC connection and how much bandwidth with the ethernet connection?
> > >
> > >
> > > regards,
> > >
> > >
> > >
> > >
>
>
>
-- http://www.freelists.org/webpage/oracle-lReceived on Tue May 05 2015 - 20:54:10 CEST