I did a thing, have a quick Linux question
-
I'd do RAID1, or RAID 6... I've only got ~3TB of data, but only 2 x 3TB drives (one of them is my backup drive at the moment).
If I don't have a real RAID controller, I'd use mdadm for Linux. I've used it in the past, and it worked very well.
-
@Sparkum said in I did a thing, have a quick Linux question:
And maybe this is just me going from Windows to Linux, I admittedly don't know anything about how harddrives work in Linux
The thing to know is that software raid is totally unreliable in windows and very reliable in linux
-
And FakeRAID in linux will (every time) show you all of the drives. It will not present a single disk to you. It will show all of the disks in the "array" as individual disks. Because FakeRAID is dangerous and linux makes that very clear.
-
I'm making an assumption right now because I think I pretty much understand the way pooling works in relation to HD IOPS and I'm highly doubting you get any of the real benefit of a raid doing it that way--at least speed wise. Hypothetical scenario:
You create a software raid in ZFS with 4 hard drives in pool1. let's say 1200 IOPS total for this pool.
Later you add 4 hard drives to that raid but it's added in pool2. Each pool is 1200 IOPS, not 2400 IOPS.FakeRAID is probably non-existent IOPS gains. It's like grouping for the sake of a label (I think). @JaredBusch could explain this better than I could though.
-
@wirestyle22 said in I did a thing, have a quick Linux question:
FakeRAID is probably non-existent IOPS gains. It's like grouping for the sake of a label (I think). @JaredBusch could explain this better than I could though.
FakeRAID has all the IOPS gains and the mirroring or redundancy. The Fake refers to the fact that it is built to trick you into think that it is hardware, when it is not. The RAID portion is real.
-
@scottalanmiller said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
FakeRAID is probably non-existent IOPS gains. It's like grouping for the sake of a label (I think). @JaredBusch could explain this better than I could though.
FakeRAID has all the IOPS gains and the mirroring or redundancy. The Fake refers to the fact that it is built to trick you into think that it is hardware, when it is not. The RAID portion is real.
Glad to be wrong about the raid portion of it but you can only Raid 0 or 1 with nothing nested and only whole disks. No hot spares and no hot swappable drives. I don't see why anyone would do it.
-
@wirestyle22 said in I did a thing, have a quick Linux question:
@scottalanmiller said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
FakeRAID is probably non-existent IOPS gains. It's like grouping for the sake of a label (I think). @JaredBusch could explain this better than I could though.
FakeRAID has all the IOPS gains and the mirroring or redundancy. The Fake refers to the fact that it is built to trick you into think that it is hardware, when it is not. The RAID portion is real.
Glad to be wrong but you can only Raid 0 or 1 with nothing nested and only whole disks. No hot spares and no hot swappable drives. I don't see why anyone would do it.
Raid 0, I see very very very few cases where it'd be useful... But on a real raid controller, you can do hot spares for raid 1.
-
@dafyre said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
@scottalanmiller said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
FakeRAID is probably non-existent IOPS gains. It's like grouping for the sake of a label (I think). @JaredBusch could explain this better than I could though.
FakeRAID has all the IOPS gains and the mirroring or redundancy. The Fake refers to the fact that it is built to trick you into think that it is hardware, when it is not. The RAID portion is real.
Glad to be wrong but you can only Raid 0 or 1 with nothing nested and only whole disks. No hot spares and no hot swappable drives. I don't see why anyone would do it.
Raid 0, I see very very very few cases where it'd be useful... But on a real raid controller, you can do hot spares for raid 1.
On a real controller yeah but we are talking about FakeRAID.
-
@wirestyle22 said in I did a thing, have a quick Linux question:
@dafyre said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
@scottalanmiller said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
FakeRAID is probably non-existent IOPS gains. It's like grouping for the sake of a label (I think). @JaredBusch could explain this better than I could though.
FakeRAID has all the IOPS gains and the mirroring or redundancy. The Fake refers to the fact that it is built to trick you into think that it is hardware, when it is not. The RAID portion is real.
Glad to be wrong but you can only Raid 0 or 1 with nothing nested and only whole disks. No hot spares and no hot swappable drives. I don't see why anyone would do it.
Raid 0, I see very very very few cases where it'd be useful... But on a real raid controller, you can do hot spares for raid 1.
On a real controller yeah but we are talking about FakeRAID.
True. The need for a hot spare is not quite as critical in RAID 1 as it would be in RAID 5 or 6... Can the fakeRAID controllers do anything other than 1 and maybe 5?
-
@DustinB3403 said in I did a thing, have a quick Linux question:
@Sparkum said in I did a thing, have a quick Linux question:
Are you able to group harddrives in a non raid format with linux?
Like a stablebit drive pool for linux kind of thing?
Versus making raid 0
Why would you do this, when you could use MD Raid and have a highly resilient solution?
Because if the drives are different sizes then you are limited to the size of the smallest drive in your array. With ZFS or Btrfs you can have different sized drives and the data span across all of them.
-
@dafyre said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
@scottalanmiller said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
FakeRAID is probably non-existent IOPS gains. It's like grouping for the sake of a label (I think). @JaredBusch could explain this better than I could though.
FakeRAID has all the IOPS gains and the mirroring or redundancy. The Fake refers to the fact that it is built to trick you into think that it is hardware, when it is not. The RAID portion is real.
Glad to be wrong but you can only Raid 0 or 1 with nothing nested and only whole disks. No hot spares and no hot swappable drives. I don't see why anyone would do it.
Raid 0, I see very very very few cases where it'd be useful... But on a real raid controller, you can do hot spares for raid 1.
SSD caching benefits greatly from Raid 0
-
@wirestyle22 said in I did a thing, have a quick Linux question:
@dafyre said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
@scottalanmiller said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
FakeRAID is probably non-existent IOPS gains. It's like grouping for the sake of a label (I think). @JaredBusch could explain this better than I could though.
FakeRAID has all the IOPS gains and the mirroring or redundancy. The Fake refers to the fact that it is built to trick you into think that it is hardware, when it is not. The RAID portion is real.
Glad to be wrong but you can only Raid 0 or 1 with nothing nested and only whole disks. No hot spares and no hot swappable drives. I don't see why anyone would do it.
Raid 0, I see very very very few cases where it'd be useful... But on a real raid controller, you can do hot spares for raid 1.
SSD caching benefits greatly from Raid 0
That falls under one of the few use cases where RAID 0 would be useful, lol.
-
@stacksofplates said in I did a thing, have a quick Linux question:
@DustinB3403 said in I did a thing, have a quick Linux question:
@Sparkum said in I did a thing, have a quick Linux question:
Are you able to group harddrives in a non raid format with linux?
Like a stablebit drive pool for linux kind of thing?
Versus making raid 0
Why would you do this, when you could use MD Raid and have a highly resilient solution?
Because if the drives are different sizes then you are limited to the size of the smallest drive in your array. With ZFS or Btrfs you can have different sized drives and the data span across all of them.
100% this.
-
@dafyre said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
@dafyre said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
@scottalanmiller said in I did a thing, have a quick Linux question:
@wirestyle22 said in I did a thing, have a quick Linux question:
FakeRAID is probably non-existent IOPS gains. It's like grouping for the sake of a label (I think). @JaredBusch could explain this better than I could though.
FakeRAID has all the IOPS gains and the mirroring or redundancy. The Fake refers to the fact that it is built to trick you into think that it is hardware, when it is not. The RAID portion is real.
Glad to be wrong but you can only Raid 0 or 1 with nothing nested and only whole disks. No hot spares and no hot swappable drives. I don't see why anyone would do it.
Raid 0, I see very very very few cases where it'd be useful... But on a real raid controller, you can do hot spares for raid 1.
SSD caching benefits greatly from Raid 0
That falls under one of the few use cases where RAID 0 would be useful, lol.
As a component of raid 10? You could put a bunch of drives together to delete everything on them faster? lol
-
22TB ? mdadm + RAID6...
-
Do all three.
Set yourself a challenge that across a 12 month period, you will learn all 3 environments. Then you'll be far better served across any job role, you'll understand the strengths and weaknesses of them and how to mange them.
The biggest career set back for any IT guy is complacency and standing still, keep pushing yourself out of your comfort zone.
And whilst you are tearing 1 environment down, you get to learn how to migrate between hyper-visors as well.
-
@Breffni-Potter said in I did a thing, have a quick Linux question:
Do all three.
Set yourself a challenge that across a 12 month period, you will learn all 3 environments. Then you'll be far better served across any job role, you'll understand the strengths and weaknesses of them and how to mange them.
The biggest career set back for any IT guy is complacency and standing still, keep pushing yourself out of your comfort zone.
And whilst you are tearing 1 environment down, you get to learn how to migrate between hyper-visors as well.
This all plays into the "why a home lab" thread.
-
@r0dISK said in I did a thing, have a quick Linux question:
22TB ? mdadm + RAID6...
That's a lot to have under RAID 6, but for a home lab is fine.
