HA CLuster PVE 1.7

[LordOmega]

Hi to all, and TKS alot for the reply at this post.
As the title shows, i need to set-up a HA Cluster.
We have:
ONE Workstation with 2 raid-1 Sata HDD only for PVE 1.7 system and template/iso storage QUAD Core 8Gb Ram 1Gbit

One Workstation with 2 SAS HDD only for PVE system too Raid 1 Double dual core xeon 16Gb ram 1Gbit with PVE 1.7 too

ONE hp proliand dl380 with 6 hdd RAID 4HDD( raid 5 ) 2HDD(raid-1) with openfiler for iSCSI Protocol and so make a SAN

and a switch Gbit cisco too.

The question is
I have clusterized the 2 workstation as PVE CLuster from PVECTA ssh command, 1 Master and 1 Slave, the KVM Vmachines work perfectly under the 2 nodes of the cluster, having the HDD into the SAN with PVE iSCSI Target as storage.

At this point: how can i make possible that if one workstation will go down, the VMs will be migrate they was into the broken Workstation? explaining: if into one VM i have my excel Document and i didn't save it before the Workstation fault, how can i make that the Vm will go up automatically with the ONLINE data not saved?

TKS Alot!!!

 

[udo]

Hi to all, and TKS alot for the reply at this post.
As the title shows, i need to set-up a HA Cluster.
We have:
ONE Workstation with 2 raid-1 Sata HDD only for PVE 1.7 system and template/iso storage QUAD Core 8Gb Ram 1Gbit

One Workstation with 2 SAS HDD only for PVE system too Raid 1 Double dual core xeon 16Gb ram 1Gbit with PVE 1.7 too

ONE hp proliand dl380 with 6 hdd RAID 4HDD( raid 5 ) 2HDD(raid-1) with openfiler for iSCSI Protocol and so make a SAN

and a switch Gbit cisco too.

The question is
I have clusterized the 2 workstation as PVE CLuster from PVECTA ssh command, 1 Master and 1 Slave, the KVM Vmachines work perfectly under the 2 nodes of the cluster, having the HDD into the SAN with PVE iSCSI Target as storage.

At this point: how can i make possible that if one workstation will go down, the VMs will be migrate they was into the broken Workstation? explaining: if into one VM i have my excel Document and i didn't save it before the Workstation fault, how can i make that the Vm will go up automatically with the ONLINE data not saved?

TKS Alot!!!

Hi,
if the node with the running VM goes away, your open data inside the VM do the same. It's depends on the Program which you use. OpenOffice write often in a tempfile, from which you can recover the latest avaible version.

To start the VM automaticly:
you must store the vm-configs (/etc/qemu-server/VMID.conf) on the other server (like /backup/node1/qemu-server/110.conf ). If node1 go down, a script can copy the conf to /etc/qemu-server and start the vm (qm start 110).


Udo

 

[jekader]

I'd also be interested in doing a HA cluster. If the nodes have shared storage, the HDD contents on the guests should be sync'ed. But for HA as you've described it, it would be necessary to somhow sync the RAM contents all the time. I don't think that is currently possible.

The thing that comes to my mind and is easy to implement - is an auto-backup mechanism and some daemon, sitting on a node and monitoring the hosts. If one goes down, the daemon just restores the latest backup of it's VMs on another nodes and starts them. Of course, all the changes since the backup will be lost, but the system will still be available. And with shared storage, even less information will get list - only the stuff still in RAM.

 

[snowman66]

Why don't you use DBRD?