Hello,

 

I have a shiny new 12 HDD server ready to move my setup to. Currently I'm running a 6 drive HP Microserver.

 

I don't want to purchase a completely new licence as part of the upgrade path if possible. Trying to keep costs down.

 

The new drives are all 3TB in comparison to 2TB drives in the existing setup.

 

Would the best way to do this be to: -

 

• Remove six drives from the new server, replace with the existing six 2TB drives in the correct order.
  
• Boot the server in this config
  
• Upgrade the licence to Pro
  
• Upgrade the parity drive to one of the new 3TB drives. (Do I just swap the drive, or allocate the parity to on of the other six 3TB drives in the machine?)
  
• Add the other mounted 3 TB drives drives to array
  

 

This is where I'm a bit lost.... How do I then move the data from the 2TB drives to the 3TB drives and replace the 2TB drives with the other 3TB ones?

 

 

Confused!

What version of unRAID? Version 5 will figure out the correct order from the serial numbers.

 

You could just start with only the old drives in the new system. Then upgrade parity with one of the new drives since it must be as large or larger than any others. unRAID will build the new parity drive from the contents of the old data drives.

 

Then upgrade each of the old disks one at a time. unRAID will take care of rebuilding the old data onto each new disk and you don't have to copy anything.

 

Of course, preclear each disk before you add it to the array.

 

There are probably some other ways that might be faster but they would be a lot more complicated.

 

There's no need to stop using the current disks. The new system will recognize them as data disks. Assign one of the new drives as parity and the all of the other drives as data. The old parity drive and the new data disks will need to be cleared. Use pre-clear for best results.

dgaschk

 

I have 12 3TB drives to fill the new server so I don't want to use the old 2TB drives anymore. They will be sold with the Microserver, hence the copying bit.

 

 

Do you already have these drives? Do you actually need all of the additional space right now? Since your trying to keep costs down, buying the disks one at a time as needed will be much less expensive. How many TB do you anticipate writing each month? In a few months 4TB drives will be cheaper than 3TB drives are currently. UnRAID allows mixed disk sizes and incremental storage expansion; this is one of its great features. Additionally, spacing drive purchases out over time avoids multiple concurrent failures due to a bad batch of drives.

No-one said it yet, but before you start to do anything like changing the parity drive or any data drives, you should really run a non-correcting parity check to ensure that all data from all drives can be read without error.  Only then look at changing out the parity drive to a larger one.  While doing that, do not add any new files to the server until after the new parity drive is in place with good parity data that has then been checked with another non-correcting check. Throughout all of that, keep the old parity drive to hand, just in case you need to go back to it.

I do already have the drives as they came with the server which turned up today from SGI Rackmount Systems. 12 3TB Constellation ES.2 drives included :-)

 

Looks like a bit of a plan is forming: -

 

• Run parity check on the existing array. (Running now)
  
• Boot new server with existing USB and existing drives, no new drives in place
  
• Use existing server to pre clear new drives, 2 or 3 at a time (To be verified)
  
• Swap out the parity drive for a new pre cleared 3TB drive
  
• Rebuild the array
  
• Swap one drive at a time and rebuild the array each time (To be verified)
  
• Upgrade to Pro licence
  
• Add new drives (Pre cleared or not?)
  

 

And corrections or additions would be welcomed.

 

Regards,

 

Simpic

Preclearing all NEW drives (preferably using three cycles) will get the drives past the earlier failure mode point (Google 'bathtub curve') where the drive is most likely to fail.  The worst thing to have have happen is to load an array up with new drives and then have two (or, even worst more than two) fail in a short period such that you can't replace the first failed drive before the second one happens.  Some people have reported failure rates on new drives as high as 20% in some cases.  (And to be fair, other people buy ten drives at one time and never have a problem.)  But always be aware the  Murphy laws will crop up when one trusts to good luck. 

 

In cases where you have purchased a number of drives, preclearing will (hopefully) weed out those early failures when their correction is much simpler (RMA the drive) than going through the hassle of attempting to recover/rebuilt the data on a bad drive. 

The simplest approach is to just initialize your new server with a new Pro license; then just copy all the data from the old server to the new.

 

This will cost you a new Pro license -- but you indicated you need to buy a Pro upgrade anyway (and it certainly wouldn't hurt to take advantage of the "2-pack" pricing and get a spare license for future use).

 

The cost is very minor ... especially considering what you've got invested in the new server and 12 ES.2 drives 

... by the way, I absolutely agree with Frank's suggestion that you pre-clear all the drives NOW while they're still "RMA'able".    You can do 6-at-a-time by toggling between Linux console sessions with Alt-F1, ..2, ..3, etc. and starting 6 instances of pre-clear.    For 3TB drives with multiple passes, this will take 2-3 days for each group of 6.    But then you'll have very high confidence in the integrity of the array ... and, more importantly, if you have any infant-mortality on the drives you'll have identified that BEFORE you actually load your data on them.