Growing a ZFS raidZ pool

danzi · Post by **danzi** » 03 Oct 2016 14:33

Ok, here's the deal.

3x1.5 Tb disks in a raidZ1 config. I wish to grow it. I know that one could replace drives one by one and re-silver every time a disk gets replaced.

My idea is to buy one disk at a time - I know this would only grow the pool once I get the last 1.5 Tb drive replaced... however, I am thinking of buying drives say every 6 month. Here's the pattern:

1.5 1.5 1.5 = useable space ~3 Tb
1.5 1.5 2.0 = useable space ~3 Tb
1.5 3.0 2.0 = useable space ~3 Tb
4.0 3.0 2.0 = useable space ~4 Tb

and this way gradually grow every time replacing the smallest disk with the best value... mind you at the time I think the 3Tb Red NAS is the best value for money. But I do not plan to buy 3 of them all at once.

Any pros and cons regarding the above?

Makes sense? Or stupid thing?

Thanks
D.

mooblie · Post by **mooblie** » 03 Oct 2016 20:06

I believe I read that the 3TB WD Reds had some reliability issues, that were NOT present in the 1, 2 and 4 TB models. I would go for 4TB drives as replacements if I were you (IMHO!) - especially as the price will fall over the 18 months you plan to buy them.

In my own case, I bought 4 x used 1TB WD Red drives at about £30 GBP each, to replace a mixed selection of 4 x 0.5TB drives - worked a treat! and cheap too! SMART reports no errors at all.

(BTW: I use an HP N40L - the arch-rival to the Dell T20 !! - although I see in another thread about slow transfers, you have a Xeon CPU - very envious !!! )

danzi · Post by **danzi** » 04 Oct 2016 19:21

mooblie wrote: (BTW: I use an HP N40L - the arch-rival to the Dell T20 !! - although I see in another thread about slow transfers, you have a Xeon CPU - very envious !!! )

I have az HP Micorserver N40L - this is what I replaced it with. I plan to use that for something else

But it ran reliably for 5 years now

I had version 9.3.0.x on it and did not want to mess around any longer. The T20 I bought exactly due to the CPU. I was looking at the new microserver, but due to Xeon I chose the T20.

Hm, will read on, never seen the Red NAS series to have any issues. I need to start getting new drives soon, my old ones I don't even dare to say how old they are or what brand...

mooblie · Post by **mooblie** » 05 Oct 2016 11:25

I think the survey I was thinking of is here:

https://www.backblaze.com/blog/hard-dri ... r-q2-2015/

Maybe it was Seagate 3TB to avoid!

danzi · Post by **danzi** » 05 Oct 2016 12:10

Actually I read that the 3Tb WDs have 2 platters, but only 3 heads. The 4th head is replaced with a little balance piece. This makes it assymmetrical and actually quite bad for any non-symmetric setups.

On the other hand, I was wondering, If I would go and replace these 1.5 Tb drives one by one with 3 Tb disks, I'd lose the 4k option straight away, correct? That would only work if I were to format the 3 new drives and create a new pool...

mooblie · Post by **mooblie** » 05 Oct 2016 18:45

I don't know about that. Hopefully someone else here will jump in...

garyc57 · Post by **garyc57** » 01 Nov 2016 11:57

Danzi,

Like you, I've read for YEARS that ZFS will automatically expand when adding larger drives. As you were saying...

4.0 3.0 2.0 = useable space ~4 Tb

when that third drive gets replaced with a larger drive, the pool will become larger.

However, the other day, I was playing with a test server, trying this very concept. When I offline-ed the smallest drive, and replaced it with a larger one, then did a

Code: Select all

zpool replace tank <a big number> <drive>

I was surprised to see the pool size didn't automatically expand.

I googled the issue and discovered something I hadn't seen before. The set autoexpand command. So, I did a:

Code: Select all

zpool set autoexpand=on tank

...fully expecting the pool size to recalculate. It didn't. More googling. I then read I needed to:

Code: Select all

zpool online -e tank <drive>

TaDa! The pool size increased. I then replaced the next smaller drive with a larger drive, using the offline/online -e commands, and the pool size automatically expanded.

Now, I'm not sure if just the -e parameter is needed, or if one needs both the autoexpand AND the -e parameter, for the pool size to increase. I'll leave that as an exercise for the student.

Gary

sleid · Post by **sleid** » 02 Nov 2016 19:00

autoexpand ON BEFORE replacing the first disk, the pool size automatically expanded.

If autoexpand is OFF BEFORE replacing the first disk, online-e is necessary for the pool size to increase.

garyc57 · Post by **garyc57** » 02 Nov 2016 19:57

sleid,

Thanks for the clarification. That makes perfect sense.

Chucko · Post by **Chucko** » 06 Nov 2016 18:55

Backblaze's drive data page is here: https://www.backblaze.com/b2/hard-drive-test-data.html

This has links to all their drive reliability articles and most recent data. I bought 6 Seagate 4 TB drives based on their statistics.

XigmaNAS

Growing a ZFS raidZ pool

Growing a ZFS raidZ pool

Re: Growing a ZFS raidZ pool

Re: Growing a ZFS raidZ pool

Re: Growing a ZFS raidZ pool

Re: Growing a ZFS raidZ pool

Re: Growing a ZFS raidZ pool

Re: Growing a ZFS raidZ pool

Re: Growing a ZFS raidZ pool

Re: Growing a ZFS raidZ pool

Re: Growing a ZFS raidZ pool