ZFS Health Status and Menu Coming Soon?

[tmanok]

Hey Everyone,

I've installed OPNSense on a short-depth 1U server, I'm very impressed with everything so far, but there is one thing I'm missing. Although there is a SMART status plugin available, it does not give me any ZFS health information and I can't seem to find anything on the forum or in the web panel.

There are some really smart and in-the-know people on here @Oxygen and @danb to name a couple, hopefully there is someone who can let me know what I may have overlooked or if it is in the pipeline.

For some background, I've installed a ZFS mirror on two 500GB drives, I'm just hoping there is a config menu for checking the status/health and allowing me to easily remove and replace one of the drives. Thanks to Smartmontools in the Smart plugin I can see the model+serial of the drive to know which to replace, but I would have to jump on the CLI to issue zpool status type commands manually or enable SSH (not on my router please!).

Thanks everyone, I plan on being a frequent member here, especially while I'm still learning your awesome router OS.

[franco]

Hi there,

It's fair to say it's "coming soon" considering that since 21.7 we now have a default ZFS install option available. However, ZFS has been a long road and there were little outside contributions to make it happen sooner. So now the next goal would be to add maybe a ZFS widget to the dashboard, but there are no concrete plans or feature requests. I'm happy to push this along, but first we need to agree on a feature set for the widget and what it should not do to keep it simple and maintainable in the future.

As for configuring ZFS from the GUI for use cases such as snapshots/boot environments that might be part of a future business edition instead.


Cheers,
Franco

[dcol]

I am considering moving to ZFS, but what concerns me is the need for ECC memory by many posters on the subject. Some say not important others say you should use it. I don't mind using it but I cannot find any small PC's that support ECC. Mostly just larger server systems. I typically use fanless mini-PC's like Qotom or NUC with OPNsense since the hardware requirements aren't huge.

Does anyone know of any fanless mini-PC's that use ECC memory?

[fabian]

AFAIK ECC memory is only supported by servers and maybe workstations. You will not get that into a small fanless system.

[dcol]

Thanks, that is what I thought. I will go ahead and use ZFS without ECC.

[opnfwb]

I'd take ZFS without ECC any day over UFS without ECC, all else being equal. At least with ZFS, you get some power failure tolerance that UFS doesn't provide.

[Patrick M. Hausen]

I'd take ZFS without ECC any day over UFS without ECC, all else being equal. At least with ZFS, you get some power failure tolerance that UFS doesn't provide.

Seconded. The "scrub of death" caused by unreliable memory is a myth and has been debunked multiple times. You should have ECC in every server system. If you don't, ZFS is still the most reliable filesystem around.

https://jrs-s.net/2015/02/03/will-zfs-and-non-ecc-ram-kill-your-data/

[XeroX]

As for configuring ZFS from the GUI for use cases such as snapshots/boot environments that might be part of a future business edition instead.


Cheers,
Franco

Any plans to impelement automatic snapshots before any opnsense-code or opnsense-update? I saw this on TrueNAS.

[mimugmail]

Work should start after 22.1, now focus in testing 22.1

[abulafia]

Any plans to impelement automatic snapshots before any opnsense-code or opnsense-update? I saw this on TrueNAS.

Same mechanics (automatic boot environment creation upon pkg updates) are in Illumos/OmniOS; and probably in FreeBSD itself, too. So maybe some code / concepts can be re-used for this.

[tmanok]

Hi there,

It's fair to say it's "coming soon" considering that since 21.7 we now have a default ZFS install option available. However, ZFS has been a long road and there were little outside contributions to make it happen sooner. So now the next goal would be to add maybe a ZFS widget to the dashboard, but there are no concrete plans or feature requests. I'm happy to push this along, but first we need to agree on a feature set for the widget and what it should not do to keep it simple and maintainable in the future.

As for configuring ZFS from the GUI for use cases such as snapshots/boot environments that might be part of a future business edition instead.


Cheers,
Franco

Thanks Franco! Sorry to hear that the dev team has not had much contribution to this vital feature. As for supporting ZFS in the GUI via widget or implementing snapshots, I'd like to think that snapshots would be a vital feature so that one could role back configuration mistakes, especially in critical environments. Without such a snapshotting feature, to me it would simply make sense to run OPNSense as a VM (sad) which disregards many of the other comprehensive features in the OS meant for hardware use. Edit: It is worth mentioning the fact that OPNSense can be recovered by uploading a configuration file, I am aware of this and to be fair it probably (but may not in all cases, e.g. after updating) would recover your mistakes without requiring a reinstallation.

To summarize a few more points in my mind:

• Proxmox and TrueNAS both utilize ZFS snapshots for critical VMs / Data and have good overall ZFS GUIs
• Configurable Notifications in the GUI like TrueNAS Core or via email like Proxmox would be desirable
• Health status / overview with storage usage would be necessary.(Storage consumption exists in Lobby>Dashboard already for example)
• List of disks and the pool they belong to would be very nice.(Including capacity, Model, Serial, /dev label)
• Drive replacement would be a very very very nice feature. E.g. GUI Widget says a drive is dead (specifying serial number), easily accessible button for removing the dead drive from the pool, then adding a drive into the pool for resilvering/recovery.

Perhaps functions for the administrator such as drive replacement, snapshots, scrubs, and a task list should be somewhere else such as "System>Diagnostics" for the task log and "System>Storage>Pool Summary, Disaster Recovery, Snapshots, Scrubbing" for the rest.

I'd take ZFS without ECC any day over UFS without ECC, all else being equal. At least with ZFS, you get some power failure tolerance that UFS doesn't provide.

Seconded. The "scrub of death" caused by unreliable memory is a myth and has been debunked multiple times. You should have ECC in every server system. If you don't, ZFS is still the most reliable filesystem around.

https://jrs-s.net/2015/02/03/will-zfs-and-non-ecc-ram-kill-your-data/

Too true, I've run ZFS happily on non-server grade systems without any corruption, can't speak to UFS but I am keen to avoid it. Looking forward to the new features mentioned, sounds like I'll have to wait for 22.1 and after.

Thanks for all the responses!


Tmanok

[chrcoluk]

Dont worry about ECC, I have a lot of experience with ZFS, was an early adopter with it on FreeBSD, and I dont commonly use ECC based hardware.

Yes a lack of ECC adds an extra risk factor in terms of data corruption, but this isnt exclusive to ZFS, it can happen on any filesystem, if anything ZFS is more likely to detect it.

Also memory reliability now days isnt what it was like decades ago, keep temperatures reasonable, dont overclock and you will be fine.

Also I dont mind contributing to a snapshots feature, but my request would be if I contribute its in the normal build of opnsense.  If not I can post some how to's on how to do it via CLI.

Scrub of death rumour seems sourced from freenas/truenas users who commonly run extremely wide risky raidz arrays.