Fully Initialise/Sync SAN arrays before use?

So I've presented a very large iscsi lun/array to one of my servers and that's all visible in disk management and it's created/quick formatted the drive and looks to be ready to go - I can add test data and back it up fine.

Question on my mind is, the raid 6 array is still syncing and from previous experience it's going to take days, maybe even a week or more Thanks to the management softwares inability to give a percentage remaining time, I don't know exactly, not a big deal but..

For DAS Storage I've always waited for the full array to sync before using it but I'm wondering if I need to do the same for this SAN array or once it's presented and able to be formatted are we good to go? Best practice opinions..

Thanks

 

It'll be a bit slow while syncing, but if an array is available for access, it's available for access.

 

How big are we talking for it to take that long?

Our Hitachi AMS2300 with 2 (6+1 RAID5) arrays was done in a couple of hours. This was 14 (+1 hot spare) 15k 450GB SAS.

 

That's what I assumed 7nothing.

It's ~42TB useable.

 

if you're talking that large an array, or raw storage, wouldn't the performance impact be minimised since you'd be using such a large stripe of disks?

on a side note when the array is created I start using it, initialised or not

 

Performance not really worried about, importance on I/O is not the primary motive, pure archive data storage - I carved another smaller 1TB LUN and it was maxing gig speed so there is no problem there. Seems like everyone would just use it so I will, thanks

 

42TB usable in a single Array/LUN? Wow, what SAN/Disks/RAID level are you using if you don't mind me asking?

 

Full MD3000i+1000 of 2TB sata in RAID6. Funny thing is half is already spoken for so we'll see how long it takes to fill it haha. There is more left over but I will use that later for something else.

 

What OS are you presenting the LUN to?

It's often a much better idea to present many smaller LUNs and span over them at the OS level, particularly if you're going to have many concurrent reader/writer threads running at the OS level.

 

2008 R2. It's all on a dedicated box and switch, it will do nothing but handle this iscsi traffic so I don't think it will be an issue. The current setup is far from stressed so this is overkill. I'm only really concerned about rebuild times/rates with the LUN so big so Raid 6 is a little safe guard for now. I might try spanning another couple of smaller arrays and see if there is any difference in speed.

 

The question doesn't have a real answer (assuming you detailed array, controller and shelve design/info). Of course you can present the LUN. Will it effect performance? Probably. Will it affect performance if your IOPS are 100% read and with 100% cache hit? No.

I would spend the sync time doing a course on designing and implementing storage utilizing volume management that don't contain 48 TB LUN's.

 

Happy to hear your suggestions. Disregard the fact it's not best practice, I know this and was waiting for the probable backlash heh but this is the requirement, one large dumping ground

 

If you rely heavily on SAN snapshots, which most heavy users of SANs do, then doing this can add unnecessary complication.

 

wait what? 42tb in a single lun??? wow dude, that's nuts.

I guess a good 64bit OS can deal with this ... but this seems pretty nuts to me.

 

I know trust me but I can't see another way to do it without losing too many disks to parity while keeping this size of space useable ( setting smaller luns and spanning) as I need monster storage with the amount of disks I have and some set for reserve It's far from optimal and yes the concern is there

 

Lun's do not necessarily equal array's.
If its one large array in Raid 6 (30 disks) then you might really want to consider changing it. If your array is going to take days to rebuild, if you loose 1 disk, at any time during the rebuild period if you loose another 2 disks its all gone. I'd look at splitting that in half if you can afford the space.

 

I agree and I discussed this but it's all ok and the risks are known and everything is fine. The sync has well and truly finished believe it or not!! Very quick! So it took no time at all which is good. Without going into specifics, incomparison only a tiny portion of the data needs to be kept for good and that's being pushed to multiple locations, will cross that bridge if and when we get to it

 

Sorry, this is the nature of the beast for most SAN's.