Round 2 With FreeNAS

Status
Not open for further replies.

MrFrustrated

Cadet
Joined
Apr 3, 2017
Messages
7
Okay so my experiences with freenas to date haven't been great, and at the time I didn't have anything really representative, so.

I have now retasked a HP DL2000 (DL175e x 2Blades) 2 x 12 Disk each bay has 1TB 7200 RPM Disks, the server is currently blank.

The problem i have is I don't fully understand all the raid options available to me, but having a complete and redundant copy of data is a must, its the whole point of this exercise, the total storage capacity isn't always as important as the total IO loads, I do a fair chunk of development work and gaming, I have VMs for everything, I'd happily trade of some space for performance for the VMs.

So could someone please help me to understand the raid levels, and what options would help to best fit for me, after all the limiting factor on total performance is the network I/O and Dual GB LAN.

Thanks
 
Last edited by a moderator:

j0hnby

Explorer
Joined
Apr 22, 2017
Messages
68
Hi MrFrustrated. I agree, it's proper confusing so always worth reading and asking before diving in as it's tough to change when everything is in use. However it sounds like you are in a good position to "play" with some drive layouts, then add to, or create fresh and transfer data around as you have a lot of drives (lucky person!)

Have a look at the following for your drive configuration: https://forums.freenas.org/index.php?resources/zfs-raid-size-and-reliability-calculator.49/ - it's also available from the resources section up in the top of the page.
 

Chris Moore

Hall of Famer
Joined
May 2, 2015
Messages
10,080
I am not specifically familiar with the HP product you are working with but effectively, you have two servers inside a single chassis and each server has 12 drives attached. The speed is what you want and for that you are going to want to do the drives in mirror sets. If you use the setup wizard to configure FreeNAS, one of the steps asks how to configure your disks. At that point, all you have to do is give the pool a name and select the second option which is. "Virtualization (RAID 10)". Now, ZFS doesn't use hardware RAID but what it sets up after this is a set of striped mirrors with all the available disks. The number of disks you have should give you 6 vdevs (virtual devices) of 1TB each, minus some overhead for the filesystem and accounting for the fact that a 1TB disk is not actually 1TB of storage, you should get about 5.2 TB of usable storage, however ZFS doesn't like to be more than 80% utilized, so it is really more like 4.1 TB. It should be plenty fast though.
Does that help you get started?
Also, you will need something else to boot from. The storage pool disks can't be used for boot media AND you have to pass the disks to FreeNAS by an HBA. It creates all kinds of problems to use a hardware RAID controller and eliminates many of the benefits of using FreeNAS to begin with.
 

MrFrustrated

Cadet
Joined
Apr 3, 2017
Messages
7
Hey guys thank you both for your replys, been very illuminating ive taken the time to do some reading about the various options, RAID 10 looks good to me, i get the mirror copys i want for my data and the performance that will handle my needs.

But a few follow ups in regards to RAID 10 and disk failures, obviously in mirrors its less of a problem, but if a disk did failed, what would the immediate consequences be, would i loose the array totally untill the disk was replaced? will continue to operate just swap the disk, and what happens if a second disk failed meaning 2 were offline.

Sorry if they seem like newb questions but last time i tryed this i jumped right no no help no guidance nada, it wasn't what youd call spectacular well it was a spectacular failure
 

Chris Moore

Hall of Famer
Joined
May 2, 2015
Messages
10,080
In a mirror, one of the two drives can fail and it will continue to run as usual. If the second drive in that mirror fails, they your completely down. Unrecoverable.
In a pool of mirrors, if you completely loose a mirror, you loose the entire pool. I don't use this because I am very risk adverse. Meaning that, I avoid risk, when possible. I have, for example, two UPS units on my home server in the off chance than one of the UPS units should fail. I would prefer less performance in exchange for better guarantee that my data would survive.
You expressed a desire for speed, which is why I made the suggestion I made. There are safer ways to store data, but the speed suffers a little.
Here is an article that talks about transfer speed if you want to get more on that: https://calomel.org/zfs_raid_speed_capacity.html
Keep in mind that they are doing their testing on Linux, so it is not exactly the same as FreeNAS. This is only meant to provide general information to help you make an informed decision.
 

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,680
In a mirror, one of the two drives can fail and it will continue to run as usual. If the second drive in that mirror fails, they your completely down. Unrecoverable.
In a pool of mirrors, if you completely loose a mirror, you loose the entire pool. I don't use this because I am very risk adverse. Meaning that, I avoid risk, when possible. I have, for example, two UPS units on my home server in the off chance than one of the UPS units should fail. I would prefer less performance in exchange for better guarantee that my data would survive.

Which is why if you don't want to lose redundancy if you lose a disk in a mirror set, you don't use just two drives. You do a three-, four-, or even five-way mirror.

I only go out to three. The local design goal here is that loss of a disk should not compromise redundancy if possible.
 

Ericloewe

Server Wrangler
Moderator
Joined
Feb 15, 2014
Messages
20,194

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,680
At this point, I must grumble about asymptotes.

So, as we add drives, reliability approaches perfection. The fallacy there would be that there are multiple ways for systems to fail.
 

Stux

MVP
Joined
Jun 2, 2016
Messages
4,419
So, as we add drives, reliability approaches perfection. The fallacy there would be that there are multiple ways for systems to fail.
There comes a point, where reliability would be increased more by using the Extra resources to set up a redundant system rather than further increasing the redundancy on a single part.

Two systems, replicate one to the other.
 

MrFrustrated

Cadet
Joined
Apr 3, 2017
Messages
7
Hi guys did t want anyone to think own room the info and done a runner. Quite the opposite in fact. Testing can be hugely illuminating especially in a vm environment and simulating failures.

I have some follow up questions.

I as we know the chassis is 2 x 12 disks I have to confess I've settled for RAID 10 my logic being the odds of multiple failures whilst possible is slim, and despite my best efforts I want able to corrupt data on the array whilst simulating failures, whilst I want the impossible of all world's I was wondering if I stick to the raid 10 layout, could freenas be configured in either and active active or N+1? Whilst reliable and safe data is the biggest priority the bar minimum in terms of redundancy is going to be a must.

Again thanks to all participants so far I've learned more about nas and raid, iscsi, SATA , sas and alot in between. You input has been invaluable
 

Chris Moore

Hall of Famer
Joined
May 2, 2015
Messages
10,080
Hi guys did t want anyone to think own room the info and done a runner. Quite the opposite in fact. Testing can be hugely illuminating especially in a vm environment and simulating failures.

I have some follow up questions.

I as we know the chassis is 2 x 12 disks I have to confess I've settled for RAID 10 my logic being the odds of multiple failures whilst possible is slim, and despite my best efforts I want able to corrupt data on the array whilst simulating failures, whilst I want the impossible of all world's I was wondering if I stick to the raid 10 layout, could freenas be configured in either and active active or N+1? Whilst reliable and safe data is the biggest priority the bar minimum in terms of redundancy is going to be a must.

Again thanks to all participants so far I've learned more about nas and raid, iscsi, SATA , sas and alot in between. You input has been invaluable
You can have a 'hot' spare. Should automatically rebuild in a failure, but doesn't always work, from what I was told. I am not sure how to configure it because I don't do that.
It's also possible to do a 3way mirror but that is either a lot more disks or a lot less space.

Sent from my SAMSUNG-SGH-I537 using Tapatalk
 
Last edited:

Ericloewe

Server Wrangler
Moderator
Joined
Feb 15, 2014
Messages
20,194
I think hot spares have been working fine for a while now.
 
Status
Not open for further replies.
Top