I currently have an old array of 6x 3TB WD WD30EFRX CMR drives used for backups of VM images.
over the years I had one drive fail so I put my cold spare in production (bought 7 disks at the time)
now I have another failed drive and I am unable to source an identical drive in my country. I also cannot source its replacement, the 3TB WD30EFZX. I am not going to buy an SMR drive with ZFS.
My question is should I:
- get a WD40EFZX or WD40EFPX 4TB Drive (understanding I can only use 3TB of space on it)
- get a Seagate Ironwolf ST3000VN006 ( I hope its not 1Mb smaller in disk allocation, will that cause a problem)
- Do I buy two larger drives (18 or 20TB) to setup as mirror and replace the array.
Safest would of course be option 3 given the age of the 3TB drives.
Option 1 works, is also certainly cheaper and avoid that potential risk listed in option 2.
2 Likes
Any of those alternatives should work. If it was me, I would go with #3, but obviously that’s more expensive up front.
As for #2, if you hand out entire drives to ZFS it should leave a little bit of unallocated buffer area at the end for just the reason that replacements might be a tad smaller. Check the partition table of your drives to be sure.
E.g. here’s the partition table of one of my ZFS drives:
1007.0 KiB free space
1 12.7 TiB Solaris /usr & Mac ZFS zfs-1b94a36dxxxxxxxx
9 8.0 MiB Solaris Reserved 1
1007.5 KiB free space
1 Like
You’re asking multiple questions here at once, so let me try to knock em all out:
-
no matter what, your array needs to be healthy as quickly as possible
-
I would get n+1 of the drives you plans to use for your new array
-
insert new drive destined to be backup of new array into old array and rebuild
-
build new array and transfer
-
reclaim new drive from old array and add to new array as hot spare (or cold spare)
-
don’t forget your throughput will suffer going from more drives to fewer.