New instalaltion Drown in a sea of information

[ieronymous]

Hello

New in the Prox OS. Being reading forum and documentation, watched videos and starting from shore watching the sea I though i would be in 2-3 meters watching from where I begun with (trying to make a metaphor here ahaha). Still I m at the shore looking the same chaos as I started....
Let me tell you about the specs and the project I need iT for.
Its a Dell presicion t7600 with 64gb ram and 2 sockets with x5667 and a quadro 2000 gpu (if I can recall).It had a perc 310 which unistalled ASAP and stayed with the sas onboard intel leaving all 3 SSD's upon it intact without raid. Won an auction at ebay at a reasonable price. Since it came with one hdd WD Green 1.5TB I bought 4 SSD's one for the Os and other 3 for the VM's.
What I am trying to accomplish now:
I want to run proxmox upon which I;ll install Win Serv2012 R2(i d like 2016 but thats just what we run at work for DHCP,AD,Oracle...etc), a Win 10 Client running some kind of erp service probably I ll pass it to the server also, a Linux Distro not sure which one about yet and maybe if possible a hackintosh (in order to be able and check requests from our designers because only them have that damn machines!!!) I have already setup an OMV for shares and if things go well I d like to propose to discard 3 very old machines running the above OS and virtualize them. Later on Clustering with one more node maybe an option.Of course snashots or and backups of the VM's is mandatory also.
My questions so far has to do with the following
-Since the OS is going to be in an 128gb SSD do I have to leave only 8-16gb for it and resise the otherspace for other purposes?
-Which filesystem to use for the OS in order for trim to pass through the file system
-Upon the VM creation noticed that you just assign sockets and cpu's without knowing if the assigned ones are physical or logical cores (in Unraid there was an option to know and choose core and thread with their ID) Isnt that an option here? So there is a possibility an OS to be left with only logical cores?
-The other 3 SSD's which are going to be used for the VM's if ZFS will be their filesystem is that ok with trim (I couldnt find a relative video documentation of exactly how you accomplish it) Here I have another issue probable.... even though I unistalled the pec controller the disks are again upon the sas controller its the onboard, the one that chipset c600 of teh x79 socket has, if anone knows or has come across with something similar.
-When is the proper time to pass the pci device to the OS? Before the installation or afterwards? Didnt find the relevant field to check it
-After installation I noticed that I have a choice of the root security to be PAM or VE proxmox even though the password I set was only valid for PAN What is the difference between them?
-Last, at least for me to start that installation when I create a user coudlnt find afterwards teh field to decide if that user will be admin or whatver. I searched all permission tabs in user groups ansd so on.

Sorry in advance for making this thread so long, but I kept writing down all my questions/misunderstandings 1.5 months now and though to sign in and drop a line or two asking the experts here.

Thank you in advance!!!!

PS Feel free

 

[wolfgang]

Hi

Since the OS is going to be in an 128gb SSD do I have to leave only 8-16gb for it and resise the otherspace for other purposes?

This depence on your need.
If you use it only as rootfs then 16GB is far enough.
But if you like to store also iso and backups on it then you should have more space.
The rest of the space is default for VM/Container what lay on LVM-thin.

-Which filesystem to use for the OS in order for trim to pass through the file system

The same answer again. It defense.
In a home lab scenario I would use ext4 in a productive env I would use ZFS for the raid and the flexibility.

Upon the VM creation noticed that you just assign sockets and cpu's without knowing if the assigned ones are physical or logical cores

KVM user process and Unraid as far I know use KVM.
The only thin what you can do is pin a process to a core. This is in the most senarios a drawback. Normally the kernel is very effective to load balance the processes to the cores.
The only thing what you should do is to emulate only setting what your HW is capable of.
So do not create a VM with 14 cores make it with 2 sockets and 7 cores and enable NUMA so the VM knows about the NUMA architecture.

The other 3 SSD's which are going to be used for the VM's if ZFS will be their filesystem is that ok with trim

ZFS can not trim, but this is not necessary. All modern SSD controller does it automatically.

-When is the proper time to pass the pci device to the OS? Before the installation or afterwards?

At VM creation time.

After installation I noticed that I have a choice of the root security to be PAM or VE proxmox even though the password I set was only valid for PAN What is the difference between them?

see https://pve.proxmox.com/wiki/User_Management

 

[ieronymous]

Hey thank you for your time!!!

This depence on your need.
If you use it only as rootfs then 16GB is far enough.
But if you like to store also iso and backups on it then you should have more space.
The rest of the space is default for VM/Container what lay on LVM-thin.

So if I let the installation with its default settings will it use all the available space or like other os's will auto split /resize partition render all the other (128-20)gb as unformated/unused space?

The same answer again. It defense.
In a home lab scenario I would use ext4 in a productive env I would use ZFS for the raid and the flexibility.

Since I wont have a second ssd for the OS to mirror zfs is not so mandatory at least for the OS (except if the exchange of info/data between OS and other Apps in other Lvm's needs to be in the same filesystem - probably not)

KVM user process and Unraid as far I know use KVM.
The only thin what you can do is pin a process to a core. This is in the most senarios a drawback. Normally the kernel is very effective to load balance the processes to the cores.
The only thing what you should do is to emulate only setting what your HW is capable of.
So do not create a VM with 14 cores make it with 2 sockets and 7 cores and enable NUMA so the VM knows about the NUMA architecture.

Being pretty clear here so I ll only have to check that numa option what exactly does.

ZFS can not trim, but this is not necessary. All modern SSD controller does it automatically.

Yes but I found plenty of threads inside the forum all trying to manually run fstab command to schedule trim during the week according with that discard option (or it was only to see the avaliable space after data modification to the system because they couldn't see the space to decrease in GB's)

At VM creation time.

Heard you loud and clear on that one too

see https://pve.proxmox.com/wiki/User_Management

[/QUOTE]

I ll definitely will

Thank you once more!!

 

[wolfgang]

So if I let the installation with its default settings will it use all the available space or like other os's will auto split /resize partition render all the other (128-20)gb as unformated/unused space?

Auto split

Yes but I found plenty of threads inside the forum all trying to manually run fstab command to schedule trim during the week according with that discard option (or it was only to see the avaliable space after data modification to the system because they couldn't see the space to decrease in GB's)

This trim is for the zvols not the disks.
All thin allocated storage (images) have to be trimme0d because if you not they get full allocated.

 

[ieronymous]

Thank you
@wolfgang

Before a while i installed for second time the os on a 16gb usb stick just to check something but I came across a different situation than expected.
At first chosen the ext4 file system and by default created the swap the root=local in the webui I think and the data=local-lvm in web-ui.
Running the below afterwards

root@pve:~# lsblk
NAME               MAJ:MIN RM   SIZE RO TYPE MOUNTPOINT
sda                  8:0    0 931.5G  0 disk
└─sda1               8:1    0 931.5G  0 part
sdb                  8:16   0 931.5G  0 disk
└─sdb1               8:17   0 931.5G  0 part
sdc                  8:32   1  14.6G  0 disk
├─sdc1               8:33   1  1007K  0 part
├─sdc2               8:34   1   512M  0 part /boot/efi
└─sdc3               8:35   1  14.1G  0 part
  ├─pve-root       253:0    0   3.5G  0 lvm  /
  ├─pve-swap       253:1    0   1.8G  0 lvm  [SWAP]
  ├─pve-data_tmeta 253:2    0     1G  0 lvm
  │ └─pve-data     253:4    0   5.1G  0 lvm
  └─pve-data_tdata 253:3    0   5.1G  0 lvm
    └─pve-data     253:4    0   5.1G  0 lvm

seems that sdc3 is a partition of 14.1GB. It has 3.5GB, and 1.8GB in logical volumes, and 1GB+5.1GB in a thin-provisioned logical volume, which totals 11.4GB. This means that about 2.7G is missing.

Today i re installed the OS to the same usb stick using zfs as file system without compression without checksum (after all is not even mirror usb raid, its just a plain usb device) and after login to the webui noticed that
local its now 13,56gb (from 3.38gb was using ext4)
local-zfs (no local-lvm like in ext4) its 11,97gb (from 5,10gb was using ext4)

Does that means that if you use ext4 for the OS and zfs for the data array (upon which you wanna store the vm's themselves the iso's and containers ...etc) that there will be confusion of the new terminologies zfs adds instead of the ext4. Totally accepted if it's hard to make a meaning of my question.

Bottom line is all I wanna do is run the OS from a usb stick and somehow rellocate, map elsewhere the space of the OS uses for local and local lvm to a different than the usb-stick itslef, to a disk raid storage in order to have a point (at least for me) running seperately the OS and the VMs / stored isos ..... Isn't everyone doing this after all? But in the videos the care only to show thw interface and not the main aspect of proxmox which is first set it up correctly