Setup Cloudstack HCI

[mpbraj]

HI, Planning a test platform for installing and configuring CloudStack with 3 node KVM + storage cluster which is a HCI architecture....Also with separate storage NIC...Can someone help me with some document/Howto or some website link where I can follow...
Thanks in Advance...
Raj...

[rajujith]

@mpbraj I am not sure whether there is a document specific to HCI but you can achieve the same following the general installation guide: https://docs.cloudstack.apache.org/en/latest/installguide/index.html

[mpbraj]

I have gone through the doc, had configured a basic networking... The networking under cloudstack is so confusing ....if anyone possible to help me with how the HCI architecture physical connectivity is going to be..if the storage is separate NIC...

[rohityadavcloud]

@mpbraj my CloudStack homelab is a HCI setup with three Ubuntu 22.04 KVM hosts that are also used as storage & mgmt nodes, running on Intel NUC9 which has 1G nic used for guest/public network & 10G nic (pcie) used for mgmt/storage traffic:

Each of my hosts have 2 additional NVMes that are used as Ceph OSDs & the 3 nodes run Ceph mons etc. One of the node is used to run mgmt server & NFS secondary storage. I mainly use local storage (nvme) & Ceph in my HCI setup. Here's the cloudbr1 (10G) in my setup:

My notes are here https://rohityadav.cloud/blog/ceph/ and https://rohityadav.cloud/blog/cloudstack-kvm/

To accomplish this, all I had to do was create two bridges on each of my KVM hosts (cloudbr0 & cloudbr1), here's my netplan config for reference:

root@cms1:/home/rohit# cat /etc/netplan/00-installer-config.yaml
network:
  version: 2
  renderer: networkd
  ethernets:
    eno1:
      dhcp4: false
      dhcp6: false
      optional: true
    en112s0:
      dhcp4: false
      dhcp6: false
      optional: true
    eth0:
      dhcp4: false
      dhcp6: false
      optional: true
      match:
        macaddress: 14:02:xxxx
      set-name: eth0
      mtu: 9000
    enp1s0f1:
      dhcp4: false
      dhcp6: false
      optional: true
      match:
        macaddress: 14:02:xxx
      set-name: enp1s0f1
      mtu: 9000
  bridges:
    cloudbr0:
      addresses: [192.168.1.10/24]
      routes:
       - to: default
         via: 192.168.1.1
      nameservers:
        addresses: [192.168.1.1]
      interfaces: [eno1, en112s0]
      dhcp4: false
      dhcp6: false
      parameters:
        stp: false
        forward-delay: 0
    cloudbr1:
      addresses: [10.10.1.10/24]
      interfaces: [eth0, enp1s0f1]
      dhcp4: false
      dhcp6: false
      mtu: 9000
      parameters:
        stp: false
        forward-delay: 0

For VLAN/switching I use TP-Link Jetstream 8 Port Gigabit Smart Managed Switch which has the 1G nics of my hosts connected to it. For 10G nics, I just connected those nics on the hosts directly (i.e. I don't use a 10G switch as they are expensive) & the 10G network therefore are not routable to Internet (or my home network). When creating VLANs in the zone, i.e. for use in isolated network & VPC, I had to configure the switch for such VLANs: (my adv zone has only 6 vlans, as I mostly use shared network)

To implement the infra/networks, I use pfsense at home which implements the network firewall/routing on these VLANs:

Here's the overall topology of my setup, which is a bit outdated but gives you an overall idea:

[rohityadavcloud]

@mpbraj pl close if you're satisfied with the details you've been provided.

[mpbraj]

So can this architecture be used for KVM 3 server cluster...with approx. about 40 VM's with 2 nic in terms of overall network performance...

2ndly, whether its possible to have 2 vlan with couple of vm on first vlan(192.168.1.0/24) and some vm on the 2nd vlan(172.16.1.0/24)...

[rohityadavcloud]

Yes.

[mpbraj]

Thanks for your views and suggestions....

[mpbraj]

@hi, I am trying to build the environment in virtualbox....I have few questions...would be helpful if you are revert ...

1. In the text you have mentioned the cloudbr1 has10.10.1 series of IP which is not shown in the diagram....
2. The Mgmt server in on 192 series, while the mgmt network which is 10G has a separate network...
3. how are you connecting the 10G ports between server, while you mentioned you are not using any switch bcos its expensive..
4. what would be the IP of all the VM's

[rohityadavcloud]

3. Directly between hosts, link network is enough you don’t need a switch or router for such a network. But not recommended in production.
4. IP on VMs are based on the public network which is implemented with the 1G network.

[mpbraj]

HI...Awaiting response..

[mpbraj]

Sorry to say this, Since no response, planning to start working on XCPNG/XOA...

[mpbraj]

Ok... Is it possible to give a step by step config from scratch for the attached architecture....creating of zone, vlans, public netwok, Guest network and its IP and the external switch Vlans configuration, you can point to some url where I can refer...many of the sites I checked there is no such doc....since I am building it for the first time, there are so many things to consider including that I have been asked to migrate from vmware to some opensource platform...More or less I have understood, I might need one step by step config with one host, I will be able to take it forward... [image: image.png]

…

On Wed, Jul 24, 2024 at 9:29 AM Rohit Yadav ***@***.***> wrote: 3. Directly between hosts, link network is enough you don’t need a switch or router for such a network. But not recommended in production. 4. IP on VMs are based on the public network which is implemented with the 1G network. — Reply to this email directly, view it on GitHub <#9363 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BJYWPXTC6VQZTCPYN2NHP3TZN4RBLAVCNFSM6AAAAABKUQAXNSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTAMJTGMYDAMQ> . You are receiving this because you modified the open/close state.Message ID: ***@***.***>