[OpenIndiana-discuss] 800GB in 6 discs - 460GB in raidz

[jason matthews]

On 3/24/17 3:43 PM, Harry Putnam wrote:
> I continue to have a problem understanding the output of zfs list.

You may want zpool list, depending on what you trying to get. Let's see 
what you have done. please show us: zpool status p0

> Ok, if that is correct then it means that 6 disks when totalled
> individually adding up to 800+ GB has been reduced by nearly half to
> accomodate raidz.
Did you use raidz2 ? In any case, I almost never deploy raidz(2). 
Mirrors offer faster writes with just a minimal trade off in money and 
storage bays.

>
> This is an install on a vbox vm.  I created 6 more discs beyond 2 for
> a mirrored rpool.
The English to English translation is, you made two pools one of which 
has six disks in raidz and another pool that is a single set of mirrors
Am I following you?


[snip some stuff i could not grok]
>
> So, in round figures it loses 50 % of available space in raidz
> config.

No it shouldnt. If you used raidz2 then you would loose 1/3 of your 6 
disk pool to parity.

> I have no experience with raidz and have only ever mirrored paired
> discs.
Good man. If Jesus was a storage engineer, that is how he would do it.

>
> I put these discs in raidz in a effort to get a little more out of the
> total space.  But in fact got very very little more space.
zfs set compression=lz4 p0

>
> I guessed that I would be left with around 600 GB of space.... based
> on a wag only.

Wag?
>
> This is all assuming I haven't made some boneheaded mistake or am
> suffering from a boneheaded non-understanding of what `zfs list' tells
> us.
>
> Apparenly raidz does not really save enough space to make it worth
> doing.  That is, in a space for config sense.  Nothing really gained
> over just paired mirrors.

we'll have to review the bonehead part after you send me zpool status

Here is the only system I have in raidz. It has 24 vdevs of 4 drives. 
Each drive is 2TB.
In terms of raw storage, it looks like this:

jason at dbspare001:% echo '2*4*24' |bc
192

In terms of net storage it looks like this:
jason at dbspare001:% zpool list data
NAME   SIZE  ALLOC   FREE    CAP  DEDUP  HEALTH  ALTROOT
data   174T   111T  63.4T    63%  1.00x  ONLINE  -

I lost 18TB on 192TB raw pool or about 10% and I am using way more 
parity disks than you are. my ratio is 1:4 you claim yours is 1:6 but 
we'll see when you send zpool status p0
My 1:4 ratio is not optimal so dont copy this configuration.