Optimize boot space

[WhiteTiger]

It's true, SSD drives are cheap now; but using 2x250GB SSDs to boot is a real waste of space.

Why not plan to optimize space already at the time of installation?

Maybe creating the pool even on multiple disks and in this immediately create the space for the boot leaving the rest to create the data space after installation.

 

[danb35]

Because keeping the OS separate from the data has been a deliberate design decision since before day 1. Using the boot devices for other purposes violates this principle. See also:

I have to waste an entire drive just for booting?

This resource was originally created by user: jgreco on the TrueNAS Community Forums Archive. Please DM this account or comment in this thread to claim it. With the advent of TrueNAS Scale, we seem to be getting a large number of people from the Linux community who are not quite grasping what...

www.truenas.com

 

[sretalla]

Xigmanas (the other side of the fork that became TrueNAS) allows for it... Their support and GUI is far inferior. They have some features that work. The choice is yours.

 

[WhiteTiger]

The fact that a choice was correct in the beginning does not mean that it has to be correct even now.
I can expand my disks from 1TB up to 16TB and tomorrow even to 32TB; Instead I cannot increase the SATA ports and disk bays and I don't see a good reason to waste not only SSD space, but two SATA ports and two bays as well.

Mounting an external SSD drive to USB port, as well as being useless (a 256/512GB SSD is wasted whether you mount it internal or external) may be fine on a home NAS, but not on a professional NAS with the USB cable that flutters and an SSD attached with Velcro to the NAS case.

Other solutions have been implemented; not only Xigmanas; Openmediavault still accepts USB sticks because it loads the code into memory and does not use intensive writing on the USB stick and Synology initially loads the operating system from the motherboard and then spreads it onto the disks.
Proxmox allows partitioning on the installation disk, then the user decides how to use the extra space from the GUI.

Finally, it doesn't necessarily have to be used for data (and, IMHO, it should always be up to the user to decide); you can install tools for network and security monitoring from the GUI, improve the use of the cache, already prepare the use of L2ARC and ZIL which can then be activated from the GUI while today I should add at least two other SSDs.

If I'm not mistaken, plugins and apps are also installed on the work disks today, so the logic of program and data separation is already invalid.

In short, the answer "that's it" if you don't like change your product or fork your own is not appreciable because it is with the comparison of ideas that we can change things.

 

[sretalla]

In short, the answer "that's it" if you don't like change your product or fork your own is not appreciable because it is with the comparison that you can change things.

Or just follow one of the several "unsupported" methods supplied in the forum to do it and stop complaining.

HOWTO: setup a pair of larger SSDs for boot pool and data

Moderator Warning: This HOWTO will damage the ability of your TrueNAS appliance to properly maintain itself, will break the ability to replace failed drives, and may cause problems with upgrades and updates. TrueNAS is not designed to do this sort of partitioning, and the developers are not...

www.truenas.com

The rationale/design decision was already elaborated in the thread, so no further discussion required IMO.

The enterprise clients (for whom the product is really developed) don't even give a second thought to running a couple of $100 SSDs to boot. Time spent developing something absolutely not wanted by the primary target audience isn't very productive. There are workarounds for those who can't stand the idea of spending $30 on an SSD that you don't "use to the max".

 

[danb35]

Openmediavault still accepts USB sticks

As does TrueNAS. We don't recommend them, but nothing in TrueNAS prevents them. But you're right, other OSs do things differently. If you prefer the way they do it, perhaps you'd be better served by one of them.

 

[WhiteTiger]

I don't understand why such annoyed answers.
I did not want to offend anyone.

I am not complaining, but suggesting; nor did I mention costs.
I just talked about wasted space which could be used differently, not necessarily to save data.

 

[Etorix]

I don't understand why such annoyed answers.

Because it is a recurring topic, and the answer will not change…

It may also be pointed that there's little point in booting a home/lab NAS from a mirror. There's a point for mirrored boot drives in mission-critical systems—but then cost or "wasted" ports are irrelevant: It's just another requirement to factor in.
I also do not see the point in "recovering" some 2*0.2 TB when HDD storage is measured in tens of TB.

Get a second-hand 16 GB Optane M10 for 9.99 euros or dollars, stick it in a M.2 slot which would otherwise remain unused and be happy.

 

[danb35]

I don't understand why such annoyed answers.

Because the question comes up over, and over, and over, and over, and over again, and it seems to have ramped up in frequency lately. We're getting sick of people who think they have a brilliant new idea, never bother to look to see that it's been brought up dozens of times before, much less read any of that discussion to understand why it is the way it is (even when they're given direct links to it), and (when this history is pointed out to them) insist that this time, it's different. No, it isn't.

There are, as you observe, products that behave as you desire. TrueNAS isn't one of them, and isn't likely to be any time soon. Do with that information whatever seems best to you.

 

[Arwen]

@WhiteTiger - TrueNAS, Core or SCALE, is not the end all NAS. As has been said, it's a commercial product, so that drove it's design, (using separated boot OS verses data / apps is one design decision). The vendor simply allows free access by us. (Of course, we sort of act as a beta testers... which most of us don't mind.)

I think part of the problem is that with TrueNAS SCALE, Linux people who would have ignored a FreeBSD based NAS product, are now interested. Partly because it's "Linux", and partly because of the containerized applications. (And maybe because of the promise of clustering containerized applications...) So more attention on TrueNAS.


Their have been other people who have made comparisons between TrueNAS and other NAS products. They said this other NAS product has this feature, why can't TrueNAS?

One forum member even went off the rails and tried emotional blackmail as well as ranting, because he wanted his feature on TrueNAS. (And that TrueNAS was worthless to him and many other people without that feature.) If I remembered correctly, he want either NTFS or FAT32, (or was that ExFAT?), as a tier 1 file system on TrueNAS. Then he could move his "pool" of disks to MS-Windows in case of failure. Except TrueNAS would have to go to OS level RAID, (if RAIDing the volume), instead of ZFS. Then these disks would not be movable to MS-Windows, (if RAIDing...).

Anyway, my point is that TrueNAS is not the perfect solution for everyone.

 

[WhiteTiger]

If the question "comes up over, and over, and over, and over, and over again", it is perhaps because so many of us ask ourselves what sense does it make to use a 480GB mirror for an O.S. which is declared to be sized within 10GB.

Attention, there are not only people who use the NAS for family photos and videos and would like to use some extra space. Then, there aren't even those who can buy an M-series NAS. For example, I have to assemble a NAS for an office with 5 users.

I have never talked about costs because we are talking about a few tens of €/$ of difference between a 256GB and a 512GB.
I talked about a more efficient use of the boot disk at the system level assuming that a similar hypothesis is already being studied for the near future.
Regardless of a person's skills, I am of the opinion that one should not go and change configurations through the shell, but that everything should be done through the GUI.
In fact I mentioned, only as an example, the possibility of using the extra space for cache, zil, l2arc as preset in the GUI.

I did not intend to argue, but only to solicit a clarification on the current use of this space on the boot disk and on which future developments may be.

For your information, today I am assembling TrueNAS Cores using 7 SATA ports: 4 are for the disks, 2 for the boot and with the seventh disk there is little I can do because if I were to implement L2ARC and ZIL I would still have a mirror, unless to partition the disk from the shell. Which I am against, although I have already done so in other cases.

 

[sretalla]

I am of the opinion that one should not go and change configurations through the shell, but that everything should be done through the GUI.

A perfectly acceptable way to think.

Then, there aren't even those who can buy an M-series NAS. For example, I have to assemble a NAS for an office with 5 users.

Why wouldn't you go with the TrueNAS Mini range then? It doesn't seem like 5 users would be enough to overpower one of those.

the possibility of using the extra space for cache, zil, l2arc as preset in the GUI.

For the kind of hardware you're talking about, those things aren't likely to be a concern unless you haven't installed enough RAM (and they only make things worse if that's the case anyway).

today I am assembling TrueNAS Cores using 7 SATA ports: 4 are for the disks, 2 for the boot and with the seventh disk there is little I can do

If you insist on building your own systems with the constraints you have manufactured, I would suggest that you take advice already provided in the thread and use only a single SSD to boot which gives you 2 slots for SSD to be used for apps/jails/VMs. (or make it 6 disks for the pool)

Properly configured backups and using some cron tasks for things like emailing yourself the configs each day/week should cover most scenarios where you lose a boot pool SSD (which would be very rare).


As already stated in the thread (in slightly different words), the reason to separate the appliance OS from the rest of the functions of a NAS is clear to anyone who understands that "appliances are disposable, data is not".

It is for that reason that you are extremely unlikely to see what you're asking for arriving in any foreseeable future of this "appliance NAS OS" product.

No harm in asking, but perhaps it's time to stop.

 

[Arwen]

@WhiteTiger

As far as most of us know, iXSystems has no plans to implement dual purpose boot devices. That is what we are trying to communicate, "no future development" of that "feature request".

A clarification, the boot device should be 16GBs minimum. And if using TrueNAS SCALE, perhaps more due to future unknowns.

Another clarification, at least in the past, you could never mirror L2ARC. The "real" data is in the data pool with whatever redundancy was designed into the data pool. On loss of a L2ARC device, pool access continues uninterrupted, though potentially slower. On replacement of the failed L2ARC device, the new one will start re-populating its self automatically. You can have multiple L2ARC devices on a single pool, but this wants even more memory for the lookup table than a single L2ARC requires.

Last clarification, all ZFS pools have a ZIL in the data pool. You likely mean a SLOG, (Separate ZFS intent LOG), which does not have to be mirrored. Though obviously data safety does recommend it be so. However, NVMe drives, (Intel Optane), in M.2 slots are becoming the better choice for SLOG due to the increase in speed. So a dual M.2, or a PCIe carrier card for M.2 is the more recent recomendation for SLOG.

 

[WhiteTiger]

OK, thanks everyone for your answers.

 

[Etorix]

In fact I mentioned, only as an example, the possibility of using the extra space for cache, zil, l2arc as preset in the GUI.

That would be a real BAD IDEA™.
SLOG required Power-Loss Protection, so not your average boot drive.
L2ARC may well need high endurance, again not your average boot drive. And killing the boot drive by an overload of writes in its L2ARC second duty would not end well…

For your information, today I am assembling TrueNAS Cores using 7 SATA ports: 4 are for the disks, 2 for the boot and with the seventh disk there is little I can do because if I were to implement L2ARC and ZIL I would still have a mirror, unless to partition the disk from the shell. Which I am against, although I have already done so in other cases.

There's no need to mirror SLOG and L2ARC. L2ARC is a non-critical copy of data; upon failure one loses performance, but no data. SLOG is critical for data safety, but only in the event of an unclean shutdown; so a mirrored SLOG is to guard against the not-so-likely scenario where the NAS experience a failure AND one SLOG device does not come up upon reboot.

Since SLOG and L2ARC are about performance, these are best implemented as NVMe devices, not SATA. And only if required (mandatory sync writes for SLOG; insufficient ARC hit ratio despite high RAM for L2ARC).

 

[danb35]

As if we needed another example of why this is a bad idea:

SOLVED - Rebuild pool from drives possible? because upload configuration failed to bring the pool online

Hi, first of all, don't blame me for risking my data because the data is safely backed up. I had a Truenas Scale installed on one SSD boot disk, it was fully functional, it had and LSI HBA (IT mode) and 3 x HDDs. I've saved the config db and password. Recently the partition table of my boot SSD...

www.truenas.com

 

[WhiteTiger]

Many thanks again for your suggestions
At this point my configuration is as follows

• 1 SSD for booting
• 4xHDD 1TB for the data pool in RAIDZ-1

And I still have two SATA ports to use for 2.5 "devices (I don't have any other 3.5" HDD bays)

I don't have NVMe / M.2 ports on the motherboard

The NAS is mainly used for SMB Shared Folders, with 5 users.
The memory is 16GB and the CPU is a quad core