Design rules- FAS3020

[gabrievanzanten]

Hi

In just a very few words:

- 2 yrs ago FAS3020 was installed

- design was made by 3rd party

- demand has grown

- environment has grown from 2 heads with 2 shelfs to 6 shelfs (metro cluster)

- design has never been reviewed

and now... we have performance issues. Copying a VM from one LUN to another LUN (ESX) spikes filer cpu to 99% and latency on ESX side goes up to 200ms

Unfortunately I don't know anything about designing volumes, plexes, LUNs etc. so I'm now looking for some docs on general design rules.

I have questions like:

- how many shelfs and disks can you just "stick" together in one big volume?

- Everything is configured as Raid-DP. Is Raid-DP even the best for Oracle? On other storage I would use Raid-10 and Raid-5 for that.

- In the volume I can see a mix of 10k and 15k disks and 144GB and 300GB disks. Normally I would avoid that, can NetApp handle that? I know the WAFL system is good, but ....

Anyway... links to good docs on these topics are very welcome.

Gabrie van Zanten

http://www.GabesVirtualWorld.com

[arsenalsoccer]

How are you copying the vm? Winscp? RAID-DP is native to netapp and protects from double disk failure You should always use RAID-DP, your data is worth more than saving one or two disks from configuring a less resilient raid type. You may want to check network cables and switch saturation or firewall port blocking. Hope this helps.

[gabrievanzanten]

Hi

I'm copying using the VMware Infrastructure Client.

I understand Raid-DP offers good protection, but is it the best choice in all situations? Would you run Oracle transaction logs on Raid-DP or rather use Raid10 for example?

Gabrie

[tirtha]

NetApp offers RAID_DP or RAID4. And Ontap till date does not offer different raid for different RAID groups under the same aggregate. So if u have different container for ur transaction logs, u may have RAID4 but that would be risky!!

Check these Netapp TR's hope they may help:

http://media.netapp.com/documents/tr-3369.pdf

http://media.netapp.com/documents/wp-7009-oracle-asm.pdf

http://media.netapp.com/documents/tr-3391.pdf

[ekashpureff · Accepted solution]

Gabrie -

Welcome to NetApp Communities !

Ouch ! Sounds like you may have inherited a dino mess !

Mixing disk drive sizes is a bad thing. Mixing 10K w/ 15K disk is worse !

You can build a 16TB volume/aggregate with Data ONTAP 7.3.

Some links ...

The main NOW site doc page:

http://now.netapp.com/NOW/knowledge/docs/docs.cgi

The Data ONTAP doc page:

http://now.netapp.com/NOW/knowledge/docs/ontap/ontap_index.shtml

Database Best Practices:

http://now.netapp.com/NOW/knowledge/docs/bpg/db/oracle_faq.shtml

I hope this response has been helpful to you.

At your service,

Eugene E. Kashpureff
ekashp@kashpureff.org
Senior Systems Architect / NetApp Certified Instructor
http://www.linkedin.com/in/eugenekashpureff

(P.S. I appreciate points for helpful or correct answers.)

[gabrievanzanten]

Just a little follow up about issues that have now been pinpointed:

- Our aggregate had filled up to 97%. We didn't know this was against best practise. Now cleaned up some space and are back at 85%. Wasting about 1 TB 😞

- Turned out a number of volumes were fragmented. After running reallocate measure, two volumes reported an optimization of 6 and four volumes reported an optimization of 4. I've now scheduled optimizations to take place each saturday night. Two two oracle volumes are now back at optimization 1 and 2.

We are seeing slight improvements. Last step I'm going to take now is aligning all VMs that are on the NetApp.

Unfortunately we're running ESXi, so the NetApp mbralign tool won't work 😞 Looking for a solution to this.

Gabrie

[gabrievanzanten]

Thx, I will start with these docs 🙂