best raid group for 1tb x 12, single controller

[SUSAN8888]

Hi All,

i've a new fas2020 with single controller, and 12TB x 12. I wanted the maximum disk space with protection and i'm on data ontap 7.3.6. I've planned to create use existing aggr0 12 disks - raid dp. i will keep vol0 intact, and then vol1 for my cifs. and vol2 for my snapmirror destination (from another box).

i would need the advise of:

1) i read some where where by the maximum disk for a sata raid group is only 8 disks? is it true? how can i fully utilize my 12 disks

2) anything i miss out or could be enhanced?

3) also every time when i screwed up and i will just go to maintainance mode and initialize and wait 4 hours, is there any faster way of rebuilding the whole filer?

[scottgelb]

I tend to create separate root aggrs often...except on smaller systems it is just a big loss of space.  One aggr makes sense on this system.  The max raid group is 16 and default is 14 on SATA with 7.3.

With a 32bit aggr, you can have a max of 19 1TB data disks (2 total disks).  If you had 19 drives, we'd do something like this to have even sized raid groups...   23 total, 19 data [ (RG0 10D+2P) + (RG1 9D+2P) ].   But with 12 drives... you could do 1 spare, 9 data, 2 spare and be good to go on this system.  With 5% aggr reserve and no snapshot reserve you will get a bit over 6TB usable. 

As far as rebuilding... you can backup /etc... do a config dump of /etc... or mirror the root volume to another system.  There is always flash and netboot but to rebuild the root aggr could be time if you had a failure... probably quicker to wafliron then restore but I wouldn't create more than 1 aggr on a system this small.

[SUSAN8888]

okay, i will go ahead with raid dp with aggr0 of 12 disks. before i left the DC i have set it to initialize all disks with the option to create a flex volume. when i return to the DC all i need to do is to assign all disks to the aggr0 and i should be good to go right?

also on the rebuilding, assume i do not need any data, in case i screw up my aggr0 (added too many disks) how do i go about?

also is there any way to speeding the growing when adding disk to aggr?

many thanks!!!

[scottgelb]

make sure to leave 1 spare... then check raid.min spare and make sure set to 1 instead of 2 (should be 1 on a 2020) so it doesn't alert there aren't enough spares.

If the disks are zeroed then the grow is instant...if not zeroed then they zero which takes time.  You can pre-zero with "disk zero spares".  All disks are zeroed from the factory or rma and wouldn't be zeroed only if the disks were in an aggr before and you destroyed the aggr.

If you make a mistake and want to recreate the entire system with no data on it...you can go to special mode boot (ctrl-c to special boot menu)... then the 4a option to zero out the system and recreate a new root flexvol...then you get a zeroed new system when done waiting at the setup (hostname?) prompt... you would need to repush ontap to disk and reapply licenses.

[SUSAN8888]

Hi Scott,

thanks alot, i'm using alot of the 4a, but it's painful to wait for hours.

also for "with 5% aggr reserve and no snapshot reserve" is such setting good when i have a snapmirror turn on? what's the better practice?

[scottgelb]

I probably would change aggr reserve to 3%... just to have some space reserved if you need it later and to not fill the aggr.  For usable with no snapshot reserve that was for a best case number.  The default volume reserve is 20%... what is the best practice depends...how much retention and change rates.  It could be 10% or 40%...or less or more depending on how much space your snapshots take.  I would have some reserve for snapmirror and you would need a little sizing exercise to see how much reserve.

[SUSAN8888]

thanks agai, and for vol0, what's a good size for me to resize on a fas2020, is 20G ok?

[scottgelb]

You never want vol0 to fill up.. I'd leave it the default.. Or if you shrink it, monitor it. It contains messages and does grow from other logging. With 8.0 that wouldn't work since we need much larger but you can only run 7.3 on a 2020 so 20gb will work but I'd probably leave at 50-60gb to be safe

Typos Sent on Blackberry Wireless

[SUSAN8888]

thanks alot!! last two questions,

with your advise, of going 1 spare, then i would 2 disks for parity (raiddp), and 1 for spare. hence total disk available for usage is only 9 right?

also with aggr reserve to 3%, and vol reserve of say 20%, can i fully assign 100% to be vol1 and vol2 (snap mirror from another filer) and leaving no space for future growth (assume i do not need) or the better practice is to leave some free space on the aggr0?

[scottgelb]

Yes. 9 data. I prefer having 10% free aggr, 10% free flexvol, (and 10% free lun) at a minimum.. But size your mirror and see how much is left.

Typos Sent on Blackberry Wireless