Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations bkrike on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Mount Points on SAN using Veritas FS

Status
Not open for further replies.

NavinB

Technical User
Apr 11, 2001
277
GB
Hi,

I need some info regarding mount points(veritas FS) on SAN.
Is it better to have a no. of small mount points or just a few but large mount points?
There are around 6-8 oracle databases.So would I make 5-6 large or 15-16 small mount points ?

Any help would be greatly appreciated....

Thanx
 
Hello Navin,

From our Support team's point of view and using the df -k command it is much easier to view a system with a smaller number of mount-points. However the DBA's want separate MP's for redo's - archives - data for each database so we now have about 30, not good on the eye. It also depends on amount of SAN you have, i.e. less MP's will mean that databases can share a large amount of storage and can have a little more flexibility should they grow, but the more MP's you have means you must be more precise per allocation per MP and if a filesystem fills up how much work is involved to increase the storage, again our support favour less MP's.

Marrow




 
Thanx for the prompt response....
We have around 1.5 TB of space....each DB would be around 150GB.....So if I create around 5*200GB MP(for dbf) for 5 databases and 5*100GB MP's for (redo,ctl,other things etc)...would it be a great idea ???
 
Navin - At our site I was originally talking about a 2 x database system where we have excessive mount points for databases plus other applications and our Support team's attitude to what they saw as an inflexible system. Having thought about this a bit more, a DBA would not recommend placing all your dbf & control files for each database together on one MP because of recovery and performance, and is it OK for different databases to share mount points.
FYI these are our DBase MP's, we have six per dbase:-

/mp7/oradbase # db s/ware

DB1
/mp1/oraarch01
/mp2/oradata01
/mp3/oradata02
/mp4/oradata03
/mp5/oraredo01
/mp6/oraredo02
ditto for DB2

It would be advisable to confir with your DBA before proceeding, what do they actually recommend? I don't think I would go with your 10 MP's though.

Hope I haven't confused you too much, but there was much discussion on this subject at our site.

Marrow
 
Oracle some time ago sayed: S.A.M.E. - stripe and mirror everything
My Orcale hosts usually have less mountpoints than years ago, one Testmachine just has a /oracle/ Mountpoint and performs well.
In my eyes it makes sense to make a mountpoint from the Backup view, depending on, where the Software installation lies (backup full one a month) and where the database files lie (never backup them on filesystem basis but using eg. RMAN)

Best Regards, Franz
--
Solaris System Manager from Munich, Germany
I used to work for Sun Microsystems Support (EMEA) for 5 years
 
Thank you ...both...
So best would be to create raid 5 and have few mount point s only.....
 
RAID5 is ok, but RAID0+1 is better performance.
If you do not have enough disks R5 is ok

Best Regards, Franz
--
Solaris System Manager from Munich, Germany
I used to work for Sun Microsystems Support (EMEA) for 5 years
 
we had performance ssues on raid 5 recovery took some time due to parity check...

Oracle dont recommend putting prod on raid 5

just a comment but there are + and - 's for both

Sy UK
 
This means more confusion for me....
If I don't do S/W raid but do H/W raid...then would it be fine....???
 
Backend storage is SAN(EMC).....So I need not do H/W raid....Just will create vxfs on volumes using veritas....
Any other ideas ????
 
find out what type of configuration you have on back end..they may have RAID-1,5 OR RAID-S.You can consider striping (I think RAID-2) if they have RAID-1., in other case just do concat....and vxfs with larger blocksize..say atleast 8m...if the fs is bigger then 200gb then use even larger block size.
 
HW-RAID5 is faster than SW-RAID5 ;-)
if you have enough disks do a RAID 1 (mirror) of a RAID 0 (stripes), if you need to have an eye on the number of disks you use take Raid5 (prefered hardware raids, second choice is Software raids)

Best Regards, Franz
--
Solaris System Manager from Munich, Germany
I used to work for Sun Microsystems Support (EMEA) for 5 years
 
raid-5 is not faster , it just defetes the purpose if you create raid-5 over already raid-5'd lun.
You may end up hitting same spinle...
 
sorry I mean to say it is faster but you should not create dbl raid-5 is not faster....
 
you are right a SW-RAID5 on a HW-RAID5 is more or less the worst thing you can do... ;-)

Best Regards, Franz
--
Solaris System Manager from Munich, Germany
I used to work for Sun Microsystems Support (EMEA) for 5 years
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top