Partitioning and configuration with LVM(Logical Volume Manager)

[olimongi]

Hi, I would like to get some advice on the best way to mount and configure a proxmox with 2 HDD disks of 1TB and another of 12 TB. What would be the best way to configure it using LVM (Logical Volume Manager)?

It is for an enterprise environment where many qemu virtual machines and lxc containers will be mounted.

Thank you very much

 

[LnxBil]

It is for an enterprise environment where many qemu virtual machines and lxc containers will be mounted.

Don't use those disks and get more of the same kind.

The only useful enterprise setup would be:
- RAID1 the two 1 TB disks together for fault tolerance
- use the 12 TB disk for volatile stuff

(If you would use Software RAID / ZFS, I could also imagine partitioning the 12 TB disk in 1 TB and the rest and use 3 disks/partition of 1 TB in software RAID5, but that is suboptimal concerning the guaranteed IOPS).

 

[olimongi]

Thank you very much!

What would be ideal to create a RAID5 with 1TB disks and manage them with LVM and leave the 12TB for volatile issues?

Because in total I have 5 HDD disks and 1 SSD

2 HDD of 500GB
1 HDD of 1TB
1 HDD of 1,5TB
1 HDD of 12TB
1 120GB SSD

With all these disks how would you do it?

Don't use those disks and get more of the same kind.

The only useful enterprise setup would be:
- RAID1 the two 1 TB disks together for fault tolerance
- use the 12 TB disk for volatile stuff

(If you would use Software RAID / ZFS, I could also imagine partitioning the 12 TB disk in 1 TB and the rest and use 3 disks/partition of 1 TB in software RAID5, but that is suboptimal concerning the guaranteed IOPS).

 

[LnxBil]

With all these disks how would you do it?

I would not do it in an enterprise setup (home use could be ok if you have a lot of backups). The main goal for an enterprise is fault tolerance, which cannot be done with those disks besides space waste. Normally you'd use same sized disks and have no space (and performance) waste. I'd buy more disks of the same size and type.

So this is purely a "I don't recommend, but you can technically do this" approach:

The one SSD and the one 12 TB cannot be fault tolerant, only one of them exists. You could use the other disks and parition them so that you have 4x500GB space and could use a stripped mirror yielding 1 TB of space in which the right two disks can fail and you're golden, but any one disk can fail. The rest of the space (0,5 TB from the 1TB and 1 TB from the 1,5 TB disk) would be wasted if you want performance. With only those few disks, you will most definitely have the performance. You could also have 5 disks and do a raid5, which will give you 2 TB of usable space, but even more wasted space. If you want to use those space not to go to waste, you will have a very slow harddisk array, which is always a very bad idea.

This is absolutely useless in an enterprise setup, but I want to still mention it:

WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING
If you really don't care about fault tolerance, security and are not processing any GDPR-related data in the EU, you can also strip together everything and you will have a RAID0 over all disks, which is the fastest approach, but the most dangerous one because any disk problem will cost you ALL YOU DATA on all your disks.
WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING

So again: Go and buy the same type and size of disk, maybe a RAID controller if you want to have hardware RAID and do not want to use ZFS and use that.

 

[olimongi]

Thanks for the help and advice!!!

 

[olimongi]

I also have another server with 17 disks 2 disks are SSD 1TB and 15 SAS disks all 900GB.

In this case what would you recommend me? a RAID 1 for the operating system with the two 1TB SSD and the 15 900GB SAS disks a RAID 10? or as it would be ideal according to your experience.

This server would be the same for an enterprise environment.

Thank you very much in advance!

 

[LnxBil]

I also have another server with 17 disks 2 disks are SSD 1TB and 15 SAS disks all 900GB.

That is more like an enterprise server for me. Do you have a RAID controller in there? If so, does it have CacheCade (and a license)? If so, configure all disks in either
- 2x7 Disks in RAID10 with a spare for a total of 7x900 GB and write IOPS of 7 disks, or
- 3x5 Disks in RAID50 without a spare for a total of 12x900 GB and write IOPS of 3 disks
and use the two SSDs for a caching layer as given by your raid controller.

If you don't have a RAID controller, just an HBA, I'd go with ZFS (same RAID settings as before: stripped mirror (RAID10) or raidz1 (comparable to RAID50), I recommend stripped mirror(RAID10)) and SSD special vdev. This yields the fastest setup given your hardware and will perform well.

 

[olimongi]

Yes, I have a hp smart array p420i raid controller it does not have a CacheCade or license.

I think it would be better RAID50 to take advantage of more space and also has high fault tolerance that for the 15 disks of 900GB sas and for the two 1TB SSD make a RAID1 for storage or backups.

What do you think?

Thank you!