Best custer solution without a external storage to store vms

[Tempah]

Hello,

I want to create a Cluster with proxmox 3.1 two nodes and the vms will be store at a local device

Server1:
- 2x 72GB (Raid1)for Proxmox
- 2x 900GB (Raid1) for VMs
- 2x 1Gbit NIC, one for vmb0 (internal communication) secound for VM access


Server2:
- 2x 72GB (Raid1)for Proxmox
- 2x 900GB (Raid1) for VMs
- 2x 1Gbit NIC, one for vmb0 (internal communication) secound for VM Access



This is waht i done:

- I mount on both nodes the 900gb hdd at /srv/volume1 (ext4)
- Add the storage to proxmox each node (Storage > Directory > /srv/volume1) - On webinterface you can see it works, around 850gb free
- Create a Cluster
- Create some VMs on the local device (volume1)

But if I use the "live Migration" function (node1 vm101 > node2), I got an error "volume1" dont exist on node2. I


tried all cases:
- volume1 on both nodes (Webinterface > Storage > volume1 edit: Nodes=node1, node2)
- Shared on/off
- volumenode1 for node1 (Webinterface > Storage > edit: Nodes=node1) - volumenode2 for node2 (Webinterface > Storage > edit: Nodes=node2)

No of this Option work for me, anyone have an idea whats the best solution for using local storage in a proxmox Cluster? I know i cant use the HA function, but i want to use the live Migration function?

btw: What is GlusterFS? I think the Directory storage is not the best Option for my case.


Thank you!

yours sincerely,

Marco

 

[badji]

Bonjour,
I suggest you add a third machine for high-Availability and in each machine you add disks storage with sheepdog project or Ceph poject. Ceph is still under test : Server-ceph in pvetest.

thank's
Moula

 

[Tempah]

Thank you for your answer,

third machine? Hm.. difficult because they need to run (connection, volt etc.) and a place in the datacenter. My preferred solution would to use 2 nodes and store the VMs local on each node. I just Need the live Migration for maintenance and load balancing.

 

[thheo]

This is what I use:

http://pve.proxmox.com/wiki/Two-Node_High_Availability_Cluster

( with DRBD as shared storage )

 

[cesarpk]

( with DRBD as shared storage )

@thheo
Excuse me thheo, but drbd isn't a shared storage, is a replication storage, and this may be synchronous and/or asynchronous.

@Tempah
thheo says the right thing when say that drbd is used for local storage, if you decide to use it, you can to have HA and live migration, these two concepts come together since that 2 PVE Nodes can access to the same information of a Virtual Disk on a drbd partition that is replicated synchronously between the HDDs of PVE Nodes.

Best regards
Cesar

 

[thheo]

I didn't say it is these are naming conventions so I was just trying to point it can act like a shared storage..