HomeServer CentOS 7.3 on HP ProLiant microserver G8 G1610t

[luckypt]

Good Afternoon everyone,
I have a project (I wanted to do it a long time ago), making a HomeServer.

What I want/need in this server:
1. Save my videos, photos, documents, films, tv shows, ROMS, etc... (some critical stuff here)
2. Have my DB for Kodi to Sync with all my rpi’s at home or outside if possible (stream to 4 max devices at the same time)
3. Access to all my stuff (data) inside or outside home with any device (smartphone, tablet, PC, etc..)
4. Multi-user access (can be guests at home)
5. Server OpenVPN
6. Server IP Cam
7. Client torrent

Hardware: HP ProLiant microserver gen8 G1610T -

• • Processor: Intel Celeron 2.3 GHz
  • Ram : 4Gb de base + 1 * Kingston Technology KTH-PL316E 8Gb = total 12 GB
  • Salcar 2nd Hard Drive Caddy + DeLOCK 83877 for ODD
  • 1 SSD BX100 120GB
  • 2 * 4 TB Western Digital Red, 4TB, 6Gbps
  • 1 USB Key 4GB /boot
  • 2 free slots for backup or later add more space to the server

Configuration of Physical Disks:

USB KEY - boot (with AHCI i cant boot from SSD so i have to use microSD or USB KEY for grub)
SSD - CentOS 7.3
2* 4 TB RAID 1 - data

Install Config:

USB KEY - /boot/ EXT4
SSD : vg_centos - 120 GB XFS
/ - 20GB
home - 88 GB
swap - 4GB
2-4TB: vg_data RAID 1 XFS
/data/ - 4 TB


EDIT - Changed the config of the server, no more VM's.

I've already installed OpenVPN, Deluge.

Thanks in advance for your help and tips about this project,
LuCkY

[TrevorH]

I didn't look very closely but my first thought was that 4GB is a bit light for so many VMs.

[tunk]

If you by backup mean an 1:1-copy, then you can use rsync in e.g. a script:

Code: Select all

if /bin/mountpoint -q /mountpoint-3TB-disk
then
  /usr/bin/rsync -a --delete /data /mountpoint-3TB-disk
else
  error message, e.g. a notification e-mail  
fi

Out of curiosity, why use VMs? Why not use one CentOS installation and run everything there?

[luckypt]

I didn't look very closely but my first thought was that 4GB is a bit light for so many VMs.

I have a total of 12GB

If you by backup mean an 1:1-copy, then you can use rsync in e.g. a script:

Code: Select all

 
if /bin/mountpoint -q /mountpoint-3TB-disk
then
  /usr/bin/rsync -a --delete /data /mountpoint-3TB-disk
else
  error message, e.g. a notification e-mail  
fi

Out of curiosity, why use VMs? Why not use one CentOS installation and run everything there?

I don't like to have all in one install, but if u mean by have centOS instead of ESXI for the VM's, is what i asked cause i don't know what is better.

[hunter86_bg]

If you want so many VMs , you will need more RAM. I'd recommend you to use a single CentOS 7 installation - which will reduce the memory footprint, or use KVM machines with KSM enabled (for details : KSM)

Also, as you are planing to keep important data - consider either using a RAID1 (either LVM based or LVM ontop of MDADM) with disk caches disabled, or raid 1 (no matter LVM based or LVM ontop of mdadm) using filesystem barriers (by default ext4 and xfs do use it).

[TrevorH]

12GB should be enough I suspect, depending on the usage the VMs get. Leave a couple aside for the host and divvy the remaining 10GB up between your VMs.

[luckypt]

If you want so many VMs , you will need more RAM. I'd recommend you to use a single CentOS 7 installation - which will reduce the memory footprint, or use KVM machines with KSM enabled (for details : KSM)

Also, as you are planing to keep important data - consider either using a RAID1 (either LVM based or LVM ontop of MDADM) with disk caches disabled, or raid 1 (no matter LVM based or LVM ontop of mdadm) using filesystem barriers (by default ext4 and xfs do use it).

can u explain me how to make the partitions?
something like this? :

RAID 1 hardware on 2*4TB and make it on VM1 XFS /data without LVM?

12GB should be enough I suspect, depending on the usage the VMs get. Leave a couple aside for the host and divvy the remaining 10GB up between your VMs.

i will only make minimal instals for the vm's so 2.5 GB of ram i think its enough.
Can u tell me what do u think about using CentOS instead of ESXI for the VM's?
if i install CentOS i coud make something like this:

host - CentOS
/boot – 500MB
centosVG – 119 GB
/ - centosVG-root -> 10 GB XFS
/home 20 GB XFS
/VG - 70 GB XFS
Swap – centosVG-swap -> 4 GB
datavg 4<TB (disk2)
/data XFS

and make 1 VM for Torrent's, and 1 for IP CAM

Total 2 VM's minimal install

what do you think about that? (systems files, LVM or not, raid 1 hardware)

tks to everyone for your tips. i just want somethins stable and without to much risks to my data. (sorry about my english :p)

[TrevorH]

RAID 1 hardware on 2*4TB and make it on VM1 XFS /data without LVM?

The only problem with this is that it doesn't _have_ a hardware RAID controller, it has a B120 and that's FakeRAID and all processing is done in the drivers on the host machine. Generally speaking most linux people will recommend you run a mile from that and use the Redhat recommended way of using that controller. That's to go into the RAID controller BIOS and disable RAID and set it to run in AHCI mode and then use linux software RAID.

I'd use LVM LVs for your VM guest disk images. They are faster than files and can also be easily resized if required (well, resized larger, shrinking is more difficult).

I use CentOS on several pairs of dual socket machines with DRBD to "RAID 1" the disks over the network between the two servers. I can live migrate my VMs from one to the other easily so that I can perform maintenance on the host without incurring downtime. I use KVM for the virtualisation. I've never used vmware so cannot say anything about it.

The only drawback to xfs is that it cannot be shrunk without a backup/restore.

[hunter86_bg]

I can share my home setup.
I have 1 SSD and 4SATA disks each 500 GB 64 MB cache.
The Ssd is for the centos, while the 4 disks are in software raid.I use the software raid,as in case of MB failure - fake raid can be rebuilt on the same MB.Ontop the software raid I have setup a LVM pool via libvirt and each Logical volume is single disk for a KVM machine.

Depending on your I/O and snapshot requirements you have 2 options:
1.If you want to use internal snapshots and I/O is not important, you can create a volume group with 2 PVs (the big disks).Then you can setup 1 logical volume with a mirror option (lvm built-in) and file system ext4 (which provides write barriers and will defragment online /qcow2 causes serios fragmentation/).Then setup your VM's disks on this ext4 partion as qcow2 files (metadata preallocation is another option).
2.If you need high I/O performance or using LVM snapshots, then you can create a software raid (either raid1 or raid10 with layout far /high read, slow write speeds/).Ontop the software raid you can create LVM pool via libvirt , where each VM's disks represents a logical volume from the host's perspective.Snapshots are done from the host when the VMs are offline.

If the data is critical, consider disabling the disk cache and the cache in the VMs also.Another option is to use ext4 or XFS in order to use write barriers.If useage of older ext systems is a must - then disabling of the caches is the only option.

Edit:In your case I would consider using of KSM as you are limited of RAM.It (KSM) is numa aware,so the performance will not drop so much.

[luckypt]

Hello everyone:)

i changed the first post, i remove the VM's and now i have a simple config,
what do you thing about the config of disks?

i'm having some problems with the config of Vsftpd, i can access to my server in local/VPN but can't do it on WAN.
port forward is external 21 to internal 21 - myserverip tcp

for SSH/VPN/Webmin i changed external ports to defaults internes ports and its work great, so i don't know what am i doing wrong.

tks for your help