Proxmox HA chat

#101

What would be the best way to achieve these kind of specs over proxmox.

This means, The configurations will be VPses, but there will be a deicated server, or some dedicated servers which will make the other configurations possible

General specifications

  • High availability setup in all parts to prevent SPOF
  • High performance DB cluster with redundancy
  • Optional 2 data centers for disaster recovery with synchronization in between
  • Optional monitoring server for internal workflows and health monitoring
  • Optional testing machine for SSO testing purposes
  • Optional connection to other solution (hybrid setup)

Hardware specifications per data center

  • 2x Webserver node (8 cores, 64GB RAM, disk system 60GB)

  • 2x application node (8 cores, 48GB RAM, disk system 60GB)

  • 3x database node (4cores, 64GB RAM, disk system 60GB, disk data SSD 240GB)

  • 2x database load balancer (2 cores, 4GB RAM, disk system 60GB)

  • 2x filesystem storage node (up to 100TB of usable space, available over NFS3, ex: 32cores,
    512GB ram, disk system 60GB, disk l2arc cache SSD 2-4T, 7.200+rpm HDDs 20TB)

  • 2x WebDocuments node (8 cores, 32 GB RAM, disk system 80GB)

  • 1x testing server (2 cores, 4GB RAM, disk system 60GB)

  • 1x monitoring server (2 cores, 4GB RAM, disk system 80GB)

#102

PROXMOX VE ADMINISTRATION GUIDE - RELEASE 6.1

grafik

2 Likes
#103

i would also need to learn more about setting up a ZFS server.

#104

@oneitonitram

Hi Martin

Proxmox has ZFS and CEPH built in.

It’s easier to try to set it up on a local server, than a hosted vps, at least if you’re doing it for the first time!
You can try as often as you want, until you’re satisfied it’s working as you want, and if not, you know how to fix it…

Something with at least 32 GB RAM…

My 2 cents
Andy

#105

Andy,

I how do you make this diagrams, I mean which application do you use?

Regards,
Prashant

#106

@Prashant_Arbat

Hello Prashant and welcome to the NethServer community!

The shown diagrams are almost all taken as screenshots from my Zabbix Monitoring System, also running on NethServer…

Andy

#107

Hi Andy,

Thank you very much, Looking farward to install Zabbix now on Nethserver.

https://wiki.nethserver.org/doku.php?id=zabbix

Regards,
Prashant

3 Likes
#108

@Prashant_Arbat

Hi Prashant

Good Install, regarding Zabbix!
If you need any pointers / tips, send a PM…

Andy

#109

I am curious about routing ips.

If i have a single ip on my home servers and network.

I have multiple ips on a cloud provider.

How can i route the ips on the cloud provider, to an individual server in my home network with the one ip.

#110

Hi

You can’t route IPs which don’t “belong” to you.

IPs belong to a provider, unless you are registered with RIPE, ARIN, AFRINIC, APNIC or LACNIC.
Providers among each others do NOT use routing like normal, they use AS (Autonomous System) Routing, often BGP (Border Gateway Protocoll). If you have an AS Number, basically you CAN do real Internet Routing.

See here:


and here:

A unique ASN is allocated to each AS for use in BGP routing. ASNs are important because the ASN uniquely identifies each network on the Internet. …

AFAIK, in Europe you need at least a full class C Network to get an AS Number, besides certain requirements in schooling and costs…

Otherwise you’re just renting an IP from your Provider.

Sure you can rent a car at AVIS, but you can’t return it to a Hertz dealer! :slight_smile:

You can ask your Home Provider for more IP Adresses.

You can use internal routing to make services appear as coming from your hosted environment (eg Reverse Proxy) , but you reduce your network by 50% with this (All traffic provided to the Internet from your Home pases twice through the Hosting Provider), and it’s not the most stable and best solution. But would be possible…

My 2 cents
Andy

#111

This could be what i am trying to gamble with.

Say, i have a server athome. with say, 7 VPS.
the internet at home is 10GBPS with 1 ip address.

I get a dedicated server from a cloud provider, like OVH with their DDOS protection, then route those ip to each vps on my home server.

so that, when a specific cloud ip is queries, it lands to a specific vps on my home network.

#112

@oneitonitram

Hi Martin

Routing would not be possible - simple because TCP/IP wouldn’t allow it.

Port forwarding would work, over a VPN, with a 1:1 NAT (Not the usual 1:Many NAT).
Or Reverse proxying. Or a combination of both.

Getting a second or small block of IPs (4-8 IPs ?) from your Home Provider would be much more stable and technically better!

What you CAN use at home is Name based virtual hosting, all on the same IP. That would NOT work for certain things, like separating VoIP for 2 or more clients. But a lot of Web based stuff would work.

All depends on what you want to achieve,

My 2 cents
Andy

#113

great @Andy_Wismer i think this gives me a definitive answer to what i had in my head.

until someone, (or even me) figure out a hack around it. looks like there is always a hack for everything.

#114

Try asking your Home Provider first, how much would a small block of IPs cost (say 4, 8 or even more IPs, depending on what you need for the moment.)

Then you’ld know if your traffic migration plan “saves” money, or cost more in terms of money and headaches… :slight_smile:

If I knew exactly what you wanted connected, and where, I could plan it out for you. But if the IPs are “cheap” (Always relativ), then I’d go that route…

Or just get yourself Pritunl, create a big global network connecting all your servers (Home & Hosted) and you can route almost anything anywhere internally, or use front ends for the internet, the back-ends being at home… Around 600.- a year… :slight_smile:

My 2 cents
Andy

#115

@oneitonitram

Note that the “double jump” would create a slight latency to your services…

My 2 cents
Andy

#116

Hi,
I’m using Proxmox with three nodes as HA-Cluster.
I added VM-104 located on Node 3 to the HA-Ressource Manger.

image
image1198×348 35 KB

But constantly the VW is automatically migrated to Node 1. Even after a manual migration to Node 3, a migration back to Node 1 starts immediately.

After a manual migration to Node 2, however, this back migration does not take place automatically.
Why is it not possible to leave the VM located on Node 3?

Best regards, Marko

Addendum 1: I notice the same behavior when I manually migrate the VM-102 located on node 2 to node 3. Immediately an automatic migration to Node 1 starts.

If I then start a manual migration to Node 2, VM-102 remains there without the automatic migration to Node 1.

Addendum 2:

  • Node 1 is the Master-Node
  • Node 1, 2 and 3 are LRM
  • Group-Definition: Groupname=homecluster | retricted=no | nofailback=no | Node=Node2
#117

Hi

Can anything run on node3?
Can you install a LXC or VM on node3?

Andy

#118

Yes I can. All works fine w/o adding as HA-Ressource.

Marko

#119

@capote

Hi Marko

Did you set up fencing in Proxmox?
Even simple fencing - but it’s needed for HA…

Andy

#120

Hi Andy, I didn’t even have that on the radar. Now I activated
WATCHDOG_MODULE=iTCO_wdt in /etc/default/pve-ha-manager and rebooted all nodes.
Do I have to do anything else? The instructions in https://pve.proxmox.com/pve-docs/chapter-ha-manager.html#ha_manager_fencing are quite abstract.

thanks, Marko