Moving system dataset to root drive of freenas?

[halloween]

Hi,

i´m running freenas as a virtual machine on my esxi server. Esxi is starting from usb-stick. Virtual machines are stored on a ssd connected to the internal sata controller. I made path through of a pcie sata card to the freenas vm. On this sata controller i connected some hdds for freenas.

The system dataset is stored on my first hdd. Can i move this to the ssd on which freenas is installed? I gave the vm ssd space about 10 GB in esxi.

Is this possible or do i have to install another ssd to my sata controller?


I want to moved the system dataset because of the better spin down of my hdds. I only use it on evening, but the server is powered on 24/7.

 

[Ericloewe]

Can i move this to the ssd on which freenas is installed?

Can, yes. Good idea, no, but that applies to your whole setup, if I understand correctly.

The .system dataset must be on reliable storage (that means no single drives and definitely no USB drives), or else the system may panic if something happens to the underlying storage.

 

[halloween]

But if I move it to a raid volume, this volume will never go to sleep?

What will happen to freenas if system dataset will get lost?

If I make a raid1 for my esxi SSD, will this be better?

 

[Ericloewe]

this volume will never go to sleep?

It's been discussed to death that doing so is a bad idea.

What will happen to freenas if system dataset will get lost?

System panics, basically. Immediate admin intervention is required afterwards.

If I make a raid1 for my esxi SSD, will this be better?

Yeah, I guess...

 

[jdong]

But if I move it to a raid volume, this volume will never go to sleep?

What will happen to freenas if system dataset will get lost?

If I make a raid1 for my esxi SSD, will this be better?

Is your pool really otherwise so idle that the disks can spin down or park for a meaningful amount of time?

In my experience that's been a wild goose chase.

 

[halloween]

In freenas i have two pools. One raid1 for my private photos, documents, and so on. On single volume (only 1 hdd) mainly for recording from my sat-receiver.

I only use my sat-receiver on evening for watching tv. I very rarely use the data on my raid1 - maybe 2-3 times a week - sometimes only 1-2 times a month.


So the hdd´s would be in idle most of the time. I don´t want to power down my whole server, because i run some other virtuel machines on the esxi-ssd which must be online 24/7. But the freenas-installation is not used many times.

 

[halloween]

System panics, basically. Immediate admin intervention is required afterwards.

Can you give me a link to more information about this? Maybe a manual?

If my system panics, that would be bad, but no big problem, if i can rescue all my data without losing something. I don´t need the data within short time. If i have some downtown, that wouldn´t matter.


_______________

And if i make a backup of my freenas boot-drive, system dataset will be backed up, too. Is this a problem, if i would restore a older backup from freenas-boot drive and use my volumes or will there also be panic if system dataset is some time older (maybe 1 month)?

 

[pirateghost]

I wonder how much this horse has been beaten over the years.

Spinning down your disks does not provide enough savings to warrant the additional stress you are putting on them during spin-up/spin-down processes. It is actually more of a nuisance than anything. You will not see the savings from it.

 

[jdong]

I wonder how much this horse has been beaten over the years.

Spinning down your disks does not provide enough savings to warrant the additional stress you are putting on them during spin-up/spin-down processes. It is actually more of a nuisance than anything. You will not see the savings from it.

Thanks! I haven't read any of the previous discussions there but that's exactly the point I was trying to make. You better be really really sure that the drive will be dead idle on the order of hours before spinning up and down makes sense, and idle for at least dozens of minutes if not longer before APM disk head parking makes sense. Not only from a power savings perspective (especially spindown, since spinup disproportionately uses power and induces stress on other drives sharing the same rail as well as the PSU), but also from a disk longevity environmental perspective (producing unnecessary e-waste).

Nothing in ZFS looks engineered to optimize for reduced disk activity. Transaction groups commit on fairly frequent intervals. Synchronous writes are copied into your vdevs pretty much as soon as they're hit too. Demanding even cached metadata frequently triggers prefetching from the underlying devices. Even if you have redundancy, all block devices are proactively read in order to perform a gratuitous scrub of the requested blocks.

All of that means the faintest breath of activity near your pool will likely cause every disk in it to spin up / unpark.

 

[halloween]

I have a power meter on my server. Without harddisks i need about 20-25 Watt. With my harddisks i need about 40 Watt. So it´s double the power. The server is running 24/7 an i only have my old WD green - so this disks are not for continuos operation. This would be nop problem if i would use WD red.

And i don´t use my server so often, so that it would be useful to have the hdd´s running all the time. I only use them on evening - most of the day nobody is at home. Most of the other functions i use the server is running on ssd. Only my data is stored on the hdd´s.

 

[jdong]

We all understand drives consume power and spinning them down uses less power. That's not at all up for argument.

The argument is that ZFS in any sort of workload will tend to send small amounts of activity to every drive periodically and you won't see much spindown time.


FWIW the biggest difference between Green and Red drives is that Red drives DISABLE or hugely increase timeouts on power management features to prevent long term damage to the drive!