Hardware RAID 1 vs FreeeNAS mirror for OS

[Dotty]

Which option is best and why?
1- Installing FreeNAS OS on a RAID 1 volume (created from two SSDs using the onboard hardware RAID controller)
2- Installing FreeNAS OS on an individual SSD and later mirror it using the FreeNAS GUI

Just to be clear, Im talking about the OS install, not the data, all my data drives are wired to a JBOD HBA and I will let FreeNAS ZFS handle redundancy there.
Thanks

 

[DoThat]

Hey,

Im fairly new to FreeNAS but i think i have read that raid via raidcontrollers may cause trouble because they create a shadow partition which is only be able to delete via shell i think. but i dont have experiences about this myth.

regards

 

[mer]

ZFS mirror the SSD instead of hw raid controller.
Typically if your HW RAID controller fails, you are completely dead.
ZFS mirror works very nicely.

just my opinion

 

[2nd-in-charge]

1- Installing FreeNAS OS on a RAID 1 volume (created from two SSDs using the onboard hardware RAID controller)
2- Installing FreeNAS OS on an individual SSD and later mirror it using the FreeNAS GUI

Out of these options #2, because FreeNAS boot is ZFS and it will be able to do smart on the drives.
No need to mirror later, I'm pretty sure you can install on boot mirror.

 

[Ericloewe]

onboard hardware RAID controller)
2- Installing FreeNAS OS on an individual SSD and late

Are you sure about that? On-board sounds like Intel FakeRAID, which is little more than a driver and wouldn't work here.

 

[Dotty]

Out of these options #2, because FreeNAS boot is ZFS and it will be able to do smart on the drives.
No need to mirror later, I'm pretty sure you can install on boot mirror.

The SMART is a good point. But Im not quite sure how the mirroring on FreeNAS works on failure.
For example, lets say I install FreeNAS on one SSD and I use the GUI for mirroring (Im on 11.2. U3) , what happens if the system is up and running and the SSD fails, does FreeNAS keeps running? can I get alerts?
Next, lets assume FreeNAS failover to the surviving SSD and keeps running, what happens when I reboot the server? (because BIOS is still point to the failed SSD, so the server wont boot until I manually change BIOS to point to the surviving SSD)

 

[2nd-in-charge]

Next, lets assume FreeNAS failover to the surviving SSD and keeps running, what happens when I reboot the server? (because BIOS is still point to the failed SSD, so the server won't boot until I manually change BIOS to point to the surviving SSD)

It depends how it fails I guess. I've got two boot drives, one sitting on the on-board SATA controller and the other on the SAS controller. BIOS boot priority is to boot from the on board first, and if that fails to boot from the SAS-connected drive. I pulled the SATA-connected connected drive out and FreeNAS successfully booted from its mirror via the SAS controller. But if the primary mirror gets corrupted but not completely dead, I don't think it will boot (edit: until I change the BIOS boot order, thanks Eric :)).

what happens if the system is up and running and the SSD fails, does FreeNAS keeps running? can I get alerts?

I'm pretty sure you will. I recently offlined the SATA-connected drive (because I wanted to do SLOG test in that slot, that system is still in testing). FreeNAS kept running, because the boot-mirror was still intact as far as OS is concerned.

 

[Ericloewe]

Next, lets assume FreeNAS failover to the surviving SSD and keeps running

It doesn't "failover", it's a ZFS mirror.

What happens on boot depends on the failure, but it ranges from "boots normally" to "fails to boot until you fix the boot device order" in the setup menu.

 

[mer]

zpool scrub to trigger ZFS internal consistency checks on all devices in a vdev, periodic scrubbing and paying attention to any errors reported let you catch things early. In base FreeBSD (by implication TrueOS) you can set up zpool reporting status and smart status in the periodic scripts so you can get daily/weekly/monthly reports.

Your OS pool is likely not all that dynamic (log files are the ones changing all the time) after an install so chances for corruption are minimal
Smart can help detect failures as long as the user pays attention to results however they are reported (email, log files, etc).

Booting: both disks in the boot mirror must be marked "bootable" according to whatever scheme is being used (UEFI, BIOS). The BIOS typically will find one, then start booting from it. If it can't boot from that one, it typically goes to the next device in the boot devices list. Most people use BIOS boot list to setup "boot from this disk first, then that USB, then the DVD then PXE boot" but most times you can say "boot from the onboard SATA0 slot first, then the SAS controller, then the DVD". By setting it up that way if the onboard SATA fails to boot because the device is too corrupted, it should try the SAS controller next.

NOTE: I'm speaking in generalities, you need to stop in your BIOS and see what it allows you to do.

ETA:

What Ericloewe says :)