It looks like it ought to be possible to build a PVE cluster without using shared storage, but I'm seeing conflicting information here and in the wiki.
Is it possible to build a cluster without some sort of dedicated iSCSI or NFS storage? Having just one node export the Quorum disk seems to defeat the purpose of HA!
What I had in mind was to run a 4-node cluster using Sheepdog and IPMI fencing, but I currently have no separate shared storage. While I'm sure I could export a tiny Quorum disk off a Raspberry Pi or something equally trivial (and easily replaceable), that seems inelegant.
In the Wiki page http://pve.proxmox.com/wiki/Two-Nodes_High_Availability_Cluster, I see near the top
Does the quorum disk need to be high-performance?
On a related note, the Sheepdog page mentions that the metadata should be kept on a highly-available RAID volume; why? If I put sheepdog metadata on the OS boot volume and that disk dies, I have to reinstall PVE on a new disk and that sheepdog member is permanently dead anyway... if both the OS and a data disk die simultaneously, same result, and I rebuild. Am I missing something?
Thanks,
-Adam Thompson
athompso@athompso.net
Is it possible to build a cluster without some sort of dedicated iSCSI or NFS storage? Having just one node export the Quorum disk seems to defeat the purpose of HA!
What I had in mind was to run a 4-node cluster using Sheepdog and IPMI fencing, but I currently have no separate shared storage. While I'm sure I could export a tiny Quorum disk off a Raspberry Pi or something equally trivial (and easily replaceable), that seems inelegant.
In the Wiki page http://pve.proxmox.com/wiki/Two-Nodes_High_Availability_Cluster, I see near the top
, nowhere in the text of that page do I see any way to build without a shared quorum disk! The "workaround" appears to be using a VM inside PVE to export the shared quorum disk - am I understanding this correctly?"Although in the case of two-node clusters it is recommended to use a third, shared quorum disk partition, Proxmox allows to build the cluster without it. Let's see how."
Does the quorum disk need to be high-performance?
On a related note, the Sheepdog page mentions that the metadata should be kept on a highly-available RAID volume; why? If I put sheepdog metadata on the OS boot volume and that disk dies, I have to reinstall PVE on a new disk and that sheepdog member is permanently dead anyway... if both the OS and a data disk die simultaneously, same result, and I rebuild. Am I missing something?
Thanks,
-Adam Thompson
athompso@athompso.net