You can have a single boot drive fail completely. As long as you have a current backup of your configuration, you just install a new boot drive, install your version of TrueNAS and reload your system configuration file. You are back to 100% working. You can mirror the boot devices in TrueNAS, if you don’t think you would be able to replace the failed model but you would still need to choose the 2nd boot device in the BIOS / UEFI upon boot.
I forgot about the option to install to a ZFS setup…
Still, would it be easier to swap a failed drive and recover the system if the system were on an integrated chip, or in TrueNAS’s ZFS in either a mirror or RAID-Z3?
I’ve hot swapped failed boot drives without shutting down the system in a zfs mirror twice in the last 12 months (I used REALLY cheap drives) - I can’t imagine it getting any easier.
…I still have a config backup every now & again, just in case though.
Edit:
For the VERY paranoid, there was once a post on the old forums by the resident grinch who used a combination of zfs mirror into a raid mirror for boot drives (total of 3 redundant boot drives) & walked through how to accomplish this. Honestly overkill imo, that was like 7 nines worth of uptime paranoia levels.
I avoided that solution due to a combination of lazyness & seeing the random ‘I mixed raid with zfs & now all my data is gone, plx help’ posts
This is a bit confusing. You can only have a single or mirror of the boot pool. At least by just using the GUI. There is no GUI option to make a boot pool with any Raid-Z layouts.
Creating a Raid-Z3 VDEV would require, I think, 5 disks using the GUI.
If you want a controller card, it needs to be a HBA and not ‘RAID’. ZFS wants (needs) direct access to the disks.
What’s all the noise about HBA’s, and why can’t I use a RAID controller?
Once again, unless you’re planning for >99.999% uptime, using the built-in mirror option for your boot pool will get you where you want to be. Still, occasionally back-up the config in case of 0.0001% something.
Unless this is a build that no one will be available to touch & swap out a single failed boot drive in the mirror within the next 8 years, you’ll be golden.
That guide was well-written, but won’t fit my needs since my ITX motherboard won’t have a second slot for an IR-mode HBA.
However, my motherboard does have 8 SATA Ports: Is there any reason I can’t or shouldn’t use those as a pair of RAID 1 or 5 devices, then mirror those in the installer GUI?
Maybe your “needs” do not fit well within the contraints of a mini-ITX board?
Maybe you do not “need” redundant boot after all? The preferred way to deal with a (rare) failure of the boot device is to always keep a recent copy of the configuration file and reinstall. (Multi-Report can help with saving the configuration.) If the system is not deemed “mission critical”, that’s probably good enough.
One generally prefers to keep all SATA ports for data drives—especially with mini-ITX boards.
Otherwise, if you want to use multiple ports for boot devices go ahead, but get the terminology right: There’s no RAID 1 or RAID 5 with ZFS; you create a mirror, either during installation of afterwards, through the GUI.
I apologize, it seems I wasn’t clear: I meant I could use the motherboard’s built-in RAID functions to create two arrays in either 1 or 5, and TrueNAS could create a ZFS mirror from those two arrays.
My data drives are SAS, and will be on an IT mode SAS HBA in the ITX’s expansion slot.
The easiest way is simply to have one drive. You can use Multi_Report to automatically backup and send a copy of the config file each week so you always have a fresh copy. It really is about as simple as replace, reinstall, import the config file.
I have 2 systems each with mirrors for boot drives. It is a waste of drives and upon failure you have to tell the system to boot from the other drive. If I had to do it again it would be one drive for boot.
As far as booting off drives attached to the internal motherboard ports. Yes you can do that either a single drive or a mirror.
I don’t think you can make any kind of a raid array to boot off of even when using the internal MB ports. This holds true with both the GUI and the cli.
RAID 5 looks like a terrible choice for a boot device, which wants high IOPS and low capacity. Hardware RAID controllers are dangerous for ZFS, which is why the HA strategy then mirrors its hardware RAID 1 with a third drive: Boot from the hardware RAID1; if one drives dies, the RAID controller knows it has a valid replica to boot from; if the RAID controller corrupts ZFS structures, the mirrorred drive allow ZFS to restore.
Do the HA strategy—the third drive could be a SAS drive, or NVMe. Or drop the idea and go for a single boot device, it is likely LESS risky than a hardware RAID on its own.
Why on earth would you do this? Do you seriously need this kind of redundancy from your system? You’re talking about using 4-6 drives for your boot pool, which is ridiculously wasteful. Use one, and keep regular backups of your config file. Maybe keep a spare on hand, in case you want to be able to recover quickly from failure of the boot device.