Hi everyone, I'm new in High Availability world and I choose proxmox as my first try. I only have 2 nodes and no shared storage. I just install this 2 nodes with Proxmox 4.x and successfully create a cluster for those. My next plan is making a High Availability. I read every documentation and that lead me to create dedicated shared storage. I have an idea to mount some storage to one of my nodes and create it as an NFS. Is it possible? Or my idea is the worst idea ever? Thank you.
NFS inside Proxmox Server
- Thread starter Imilah
- Start date
Hi Imilah,
what you describe was common in the VMware with a storage appliance, that does exactly that. That works, but is not the best way to do things.
With two nodes, you should create a DRBD setup. However, HA is not possible due to the need of a third node. Yet you can still create a cluster and play around, but automatic failover is not possible.
what you describe was common in the VMware with a storage appliance, that does exactly that. That works, but is not the best way to do things.
With two nodes, you should create a DRBD setup. However, HA is not possible due to the need of a third node. Yet you can still create a cluster and play around, but automatic failover is not possible.
We use for storage dedicated server, also with proxmox. For storage in big environment we use only PVE with ZFS and 10gbit.
Hope I have one more machine to be an NSF.
So, when I create a DRBD, I will have two proxmox with redundant storage and VM, right? There will be 2 identic VM on different machine and manually I choose which one will be running, is it right? Or there is another way managing this two redundant machine
It is depending on your Apps in the VMs. It is not depending on the HostOS like Proxmox. But yes you will have 10Gbit, trust me.
Finally i create a drbd in two nodes, and downgrade my Proxmox to v3. I use direct connection from node 1 to node 2 for drbd's syncer. How about that? Do it still need a 10Gbit connection?
The 10 Gbit is only necessary if you want to have fast transfers, technically it is possible to go even lower than 1 Gbit. The interconnect limits the maximum data rate of writing on the VM, because writes are synchronously written to disk on both nodes, so you need a fast (low delay, and big throughput) connection. It'll run fine with 1 Gbit, yet not as fast as you'd like to. You'll be limited to at most 125 MB/sec. Depending on your workload and your other systems, this is maybe not a big performance killer for you (using 1 Gbit instead of 10 Gbit). Also, in case of a crash of one node, the bigger the throughput, the faster the node is completely in sync again after joining.
I'm running two nodes with DRBD over 1 Gbit and it's totally fine, yet there is no fileserver on the machines, so only some logfiles are written to disk.
I'm running two nodes with DRBD over 1 Gbit and it's totally fine, yet there is no fileserver on the machines, so only some logfiles are written to disk.
Hi LnxBil,
Hope you can share your experience on running two nodes. I have watch a video about this two nodes HA and he show a condition when the two nodes down. Let's call the first node as node A, and the second as node B. The last time before two nodes down, all VM located in node B. But only node A is able to run again. On the video I see, no way we can get the VM running on node A because before two nodes down all VM in node B. In the video said that every time we make a VM, in node A or node B, we have to backup all VM config inside /etc/PvE/qemu-server to another location in node A and node B, so we can run the VM again if the nodes never migrate the config file. Is it the only one solution to keep the VM still occur? Or you have another solution to keep the VM?