Understanding the Cluster-System

[Ray]

Hello,

I have some troubles understanding the cluster feature.

I have 3 machines:
Master 2.6.32 64bit, installed over debian, 1.5 pve
Node-1 2.6.18 64bit, installed over debian, 1.5 pve
Node-2 2.6.18 32bit, installed over debian, 1.4 pve

I have all of them in the cluster, etc. but I have problems with the storage configuration.

Why is the storage configuration syncronized?
Can I deactivate the syncronization for, say, local or backup?

My problem is that Master has enough space (> 180G) so I configured the iso space and backup as local storage.

After adding Node-2, which has only limited diskspace and is only for special purpose, the iso's got syncronized filling up node-2. And the backup storage is as well defined here, along with other storage (iscsi, lvm) which I have only defined on and for the Master.

How do I attack these problems properly, so that each node can have it's own storage asignments?

Another problem which I am not sure why it occurs is, on node-2, if I try to change any storage setting, it tells me: You have no write access!

Is this because this machine is not master or because of another (maybe installing in 32bit-mode related) problem?

On a site-node, will there be a "unofficial" repository for 1.5 as well? As I need to support my old hardware for openvz only and i would like to have it in the cluster.

thank you in advance
best
Raimund

 

[tom]

Hello,

I have some troubles understanding the cluster feature.

I have 3 machines:
Master 2.6.32 64bit, installed over debian, 1.5 pve
Node-1 2.6.18 64bit, installed over debian, 1.5 pve
Node-2 2.6.18 32bit, installed over debian, 1.4 pve

32-bit is not supported, and you need also 1.5 on all nodes.

I have all of them in the cluster, etc. but I have problems with the storage configuration.

Why is the storage configuration syncronized?

thats how it works, see http://pve.proxmox.com/wiki/Storage_Model

Can I deactivate the syncronization for, say, local or backup?

My problem is that Master has enough space (> 180G) so I configured the iso space and backup as local storage.

After adding Node-2, which has only limited diskspace and is only for special purpose, the iso's got syncronized filling up node-2.

if you have a lot of ISO, store them on a NFS or CIFS mount point. if you store them on the local storage it will be synchronized to all nodes which is waste of space.

And the backup storage is as well defined here, along with other storage (iscsi, lvm) which I have only defined on and for the Master.

all storages need to be accessible from all nodes - that's the intention.

How do I attack these problems properly, so that each node can have it's own storage asignments?

if you do not want this central management, then you should not use the cluster - run stand alone hosts as you do not want the cluster setup regarding storage.

Another problem which I am not sure why it occurs is, on node-2, if I try to change any storage setting, it tells me: You have no write access!

yes, only the master has access - described above.

Is this because this machine is not master or because of another (maybe installing in 32bit-mode related) problem?

On a site-node, will there be a "unofficial" repository for 1.5 as well? As I need to support my old hardware for openvz only and i would like to have it in the cluster.

thank you in advance
best
Raimund

a final note: the whole cluster management and behaviour is subject to change for 2.x

 

[Ray]

Thank you for your detailed answers, I will digest the wiki regarding storage model right away.

on your final note: what's the timeframe for 2.x?

best
Ray

 

[tom]

no fix schedule for 2.x, but we plan to start a longer beta phase beginning Q2/2010.