Best / quickest / least painful way to “vertically” expand a NAS when having a full data backup

---

berillio wrote:

 

But as I am constantly needing more data space, few days ago I bit the bullet and bought 2x 12TB Seagate Ironwolf, with the intention of fitting them to the the RN214a (the “wrong” NAS in that previous post), simply because adding just 2 disks would give me an extra 8TB, which should be ample for now. I could expand again at a later stage.

Has anybody used such a big array in an RN214?

 

---

I don't have an RN214, but I have two 14 TB drives in an RN202, and have not had any issues.  One is a WD Red Pro, the other is a Seagate Exos.  It should be ok.

Any particular reason you are still on 6.10.3 (2020 firmware)?

---

berillio wrote:

 

For example, I could:

1.  backup the existing configuration, then remove all the disks (if I was deleting the Volume at that moment, the disks would be wiped out and a would lose a 2nd parachute backup);
2. insert one 12TB, create the new Volume, and copy all the data;
3. insert the 2nd 12TB disk. If I have chosen X-RAID, the NAS should copy over the data making a mirror image (RAID1, I guess).
4. insert one (wiped out/reformatted) 4TB disk. The Nas should then resync going into RAID5 (I presume that at this stage, with a dissimilar size disk, the Nas would create the md126 and uses a different RAIDs for md127 and md126)
5. insert the 4th  (4TB) disk and wait for another resync.

Don't do this. Trying to add a smaller disk to the array will usually fail outright.  There are a few cases where I've seen the NAS add smaller disks when you install the second one, but you would lose capacity.

If this worked with 2x12TB+2x4TB you'd end up with 16TB instead of 20TB.  

The flaw in your reasoning here:  md127 normally is built from partitions on each disks (sda3, sdb3, sdc3, sdd3).  md126 is normally built from partitions on the two largest disks (sda4, sdb4 for example).  So for this to work properly, the two biggest disks need 4 partitions (sda1 for the OS, sda2 for swap, sda3 for md127 and sda4 for md126). sda3 is the size of the smallest disk (4 TB in this case), sda4 fills the remaining space (8 TB in this case).

But when you start with 12 TB drives, the NAS will create a 12 TB sda3 partition.  It will not attempt to repartition the 12 TB drives when you add the smaller ones.  So at best you wouldn't end up with RAID-5.  You'd end up with two RAID-1 groups instead - which wastes space.  

Also, as far as disk stress goes, it's the repeated syncs that creates the disk loads (and takes the extra time needed when upgrading all four disks).  Your process doesn't reduce the number of syncs.  If you want to start over, the way to do that is to insert all four disks and do a factory default.  That will build the volume in the most efficient way (with the least disk i/o).

I think you are over-thinking this.  In this situation I'd just do the normal vertical expansion - hotswapping one disk with the 12 TB one, waiting for the resync to complete, and then hotswapping the second one.

---

berillio wrote:

 

Could I just  do (1), then add One disk, create the Volume structure, copy half a dozen of small text file (one in each share) and the add all the disk as I said earlier? This way resync time would be a matter of seconds . When I then copy the FULL DATA, it would be striped across the array as it is copied. That should be far less demanding on the drives

 

---

RAID runs "below" the file system - creating virtual disks (md126 and md127) that the file system uses.  The resync time doesn't depend on how full the file system is - RAID neither knows that nor cares.  It is exactly the same process for an empty volume as a full one. So no time savings if you do this.

---

berillio wrote:

 

Just for info, disks age and usage is as follows:

on RN214a, 4x 4TB disks

Bay 1, 26 Jan 2020, 25205 h:

Bay 2, 15 May 2018, 34603 h;

Bay 3, 26 Jan 2020, 25213 h;

Bay 4,  7 Mar 2018, 36918 h.

          But I have a fresher 4TB disk with 18456 hours of use. I could use that one, easy done if I am rebuilding the volume from scratch

 

---

Generally when I am replacing disks for expansion, I remove the oldest.  So I'd replace disk 4 with a 12 TB model, and after the resync completes I'd replace disk 2.

If you want to use the younger spare, you could also replace the disk in either bay 1 or bay 3 with that disk.  I'd do that first, just to get it out of the way.

---

berillio wrote:

 

Incidentally, just a THERMAL concern: wouldn’t be better to place the 12TB disks in bay 1 & 4 and fit the 4TB disks in the middle, as the middle drives always tend to run a little hotter.. ? (easier done if  I choose to create the new volume from scratch, with a "conventional" expansion I would need another disk swap & resynch as the disk in Bay 2 is the the 2nd oldest disk in the existing array), 

 

---

Not sure how much it matters.  If you want to do that, you could power down and reorder the disks before you begin (swapping disk 1 and disk 2, so the oldest drives are in slots 1 and 4).  Then boot up and proceed with the replacements.

---

berillio wrote:

 

Incidentally, I was advised to “test” the new drive, before installing it, so I initially run a “quick” scan of both drives on a USB adaptor; now I am testing the first of the two drives (installed on a PC on a SATA3 socket) using Victoria 5.36 (full scan sequential test); that should take another ~7h (~14h in total to scan the entire disk).

I use Victoria because I am reasonably familiar with it, it provides easy understandable informations and it seems to work; I also remember using the no-nonsense old WDDLGD, while I am not familiar with the latest versions of SeaTools and WD dashboard.

Is there any other software or test which I should run?

 

 

---

If Victoria has a full erase test, then I'd also run that.  Seatools does have that test (dashboard does not).

My reasoning here is that I have sometimes had new disks pass the full read scan, but fail the write.  (I've also had disks that fail on the read scan, and pass on the write).