70% usage warning, but a little confused

[ThisTruenasUser]

HI
As described, there was the warning about 70% usage.
Also it states for optimal performance stay below 80%.
It seems I am apparently doing that already.
In the gui in storage, it is 76% used with:

Usable Capacity: 17.96 TiB

• Used: 13.75 TiB
• Available: 4.21 TiB

It makes little sense. I looked into it a little further:
Partial output of sudo zpool list -v
NAME SIZE ALLOC FREE CKPOINT EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
Main-Pool 18.2T 12.8T 5.43T - - 2% 70% 1.97x ONLINE /mn
dedup
mirror-1 13G 3.34G 9.66G - - 61% 25.7% - ONLINE

So that is where it seems to get the info from.

The pool is 5 x 4TB drives in raidz1 with 2 x 16GB optane for deduplication.


I am apparently making good use of dedup & compression to save lot of space:
What is the problem with the GUI? There seems little logic to the output.

Oh and for the negative nelly dedup 'haters' from zpool status -Dv:

dedup: DDT entries 4745116, size 755B on disk, 243B in core

So a quick calculation makes the 3.34GB dedup data to be spot on , with a eye watering 1GB use of memory.
The purpose in the pool is multiple sparse ISCSI shares for games, working very well.

What is the problem with the GUI? There seems little logic to the output.
What am I missing?

As for actual use of the pool, presently apparently 70%.
Why the alert at 70%?
I understand at 80% capacity, it behaves differently.

With everything I may to put on the nas, whether games or otherwise would be a maximum of 85%.
Should I even care about that?
My concern would be ISCSI performance.
Oh and not inclined to get more storage, unless a drive fails.

Thanks

 

[sretalla]

Well I'm glad you're having a good experience with dedup, you're one of the few people I've seen here in that category, but I have nothing against it being used where it works out well for everyone.

As for the warning...

If you don't want to be warned about it, go into the alert settings and say so.

The intention of the defaults is to have your awareness raised as you approach the "safe limit". If you need no such guard-rail, go ahead.

On the Alerts bell in the top right, click, then the cogwheel at the top of that menu.

Select the category: Storage and set the warnings you don't want to NEVER.

Problem solved.

 

[Etorix]

There's no hard limit. A pool won't fly at 79.9% occupancy and suddenly fall off the cliff at 80.0 or 80.1%…
A warning at 70% is not inappropriate to give time to work out a solution of expansion and burn-in new drives. But since you're doing block storage (iSCSI), the recommendation would actually be 50% or less—and mirrors, not raidz1.

You are free to decide to ignore the warning. Just be aware that if, some day, you find your system has become slow and poorly responsive, the solution will be a sizeable enlargement, preferably with a whole pool rebuild, as adding an extra vdev would not solve the condition of the existing vdev.

 

[ThisTruenasUser]

Well I'm glad you're having a good experience with dedup, you're one of the few people I've seen here in that category, but I have nothing against it being used where it works out well for everyone.

As for the warning...

If you don't want to be warned about it, go into the alert settings and say so.

The intention of the defaults is to have your awareness raised as you approach the "safe limit". If you need no such guard-rail, go ahead.

On the Alerts bell in the top right, click, then the cogwheel at the top of that menu.

Select the category: Storage and set the warnings you don't want to NEVER.

Problem solved.

I was confused about the 70% one.
Is it possible to choose the %. For me 85% would be about right.

Oh and I use 1MB block sies for the iscsi zvols. In windows format at 1MB block size. That is how I managed to get the deduplication to play nice.

 

[jgreco]

Why the alert at 70%?

To let you know. The idea is to let you know BEFORE you hit 80%. But really this should be about 40-45% if you're using iSCSI for block storage. You can survive this for a little while. Maybe even a long while if you are not doing too many overwrites. The thing that dooms block storage setups over time is the fragmentation that results from overwrites of small blocks, because both the blocks and metadata involved in those transactions are difficult for ZFS to cope with.

The pool is 5 x 4TB drives in raidz1 with 2 x 16GB optane for deduplication.

Wow, there is nothing I like in that sentence.

What is the problem with the GUI? There seems little logic to the output.

By the time you are doing block storage, you are kind of expected to realize that what the GUI is doing is really intended for typical NAS type usage, not the SAN usage that you are using the NAS for. The SAN usage ends up with a bunch of other things being important that are not emphasized in the GUI. You can still look at the NASware as an assistant for managing your storage array though.

Is it possible to choose the %. For me 85% would be about right.

Set it as you like. Storage->Pools->Edit Dataset

 

[ThisTruenasUser]

There's no hard limit. A pool won't fly at 79.9% occupancy and suddenly fall off the cliff at 80.0 or 80.1%…
A warning at 70% is not inappropriate to give time to work out a solution of expansion and burn-in new drives. But since you're doing block storage (iSCSI), the recommendation would actually be 50% or less—and mirrors, not raidz1.

You are free to decide to ignore the warning. Just be aware that if, some day, you find your system has become slow and poorly responsive, the solution will be a sizeable enlargement, preferably with a whole pool rebuild, as adding an extra vdev would not solve the condition of the existing vdev.

Really.
Very very wasteful of space as I see. It depends on the application I expect
So to buy the drives for a mirror, my setup, would have no benefit & would have no need to use deduplication,, ISCSI or even truenas.

Raidz1 is what I am using, as that is what I have.

I also use primocache software on my gaming PC. It was a masive $30US, and 1 1/2 TB drive as a caching drive, which was about the same price.
It is very effective as a block storage cache.

So to shove on the tin foil hat:
So with the general hostility to deduplication and really inefficient recommendation, could maybe suggest there are vested interests related to the computer storage industry.
I have 2 x 12TB sparse ISCSI drives for games storage,
I may go on spending spree & replace the 2 x 16GB optane with 2 x 32GB.
Then possibly create another two 12TB drives drives related linux gaming.

 

[Davvo]

May I suggest the opener to read the following resource?

Resource - The path to success for block storage

ZFS does two different things very well. One is storage of large sequentially-written files, such as archives, logs, or data files, where the file does not have the middle bits modified after creation. This is optimal for RAIDZ. It's what most...

www.truenas.com

 

[sretalla]

Is it possible to choose the %. For me 85% would be about right.

I think the question being asked there is "can I change the % used where the warning is sent?"

You have seen the screen where you can elect to not send some of the warnings and from there it's clear that the % is baked into the UI.

You can raise a bug/feature request to have it modified in some way if that's important to you.

 

[jgreco]

So with the general hostility to deduplication

The general hostility to deduplication has been a thing in the ZFS community for years. People come in expecting magic and they instead find tragic. There are certain workloads that benefit from it, especially virtual desktop OS block storage where you might have hundreds of images sharing a huge number of identical blocks. However, most users get marginal results. Part of this is because practical deduplication only happens at small block sizes (think: sector sized), but running a ZFS pool on small block sizes is incredibly inefficient. So you end up running the block size at something more ZFS-friendly which tanks the dedup efficiency.

and really inefficient recommendation,

???

could maybe suggest there are vested interests related to the computer storage industry.

Damn! I've been found out! Wonder how my storage stocks are doing... oh crap, Seagate and Western Digital are BOTH down....

But seriously, this is a compsci thing, not a nefarious plot by the two and a half remaining disk drive vendors.