I've been tasked with researching and building a new proxmox cluster to test moving away our dependancy on vmware. The main function of this cluser will be a development test bed hosting services for development and hundreds of VMs for development engineers and qa testers. I will have a 3 node cluster and trunas SAN. I would like to make this as robust and redundant as possible. I'm planning to have 6 high capacity network connections on each system, nodes and SAN. I want to setup HA, SDN, IPAM, dedicated corosync cluster traffic, and networking with OVS. Any best practices or lessons learned from others along the way.
New Proxmox Setup for Enterprise - best practices
- Thread starter kcbocca
- Start date
You are asking a lot from a single forum post.
Since we became a Proxmox Training Partner, we have noticed that our clients fall into one of two broad categories:
- Those who have taken the Proxmox VE Training.
- Those who have learned Proxmox VE on their own.
- Have fewer support issues
- The support issues they do have are more about "I would like to do this edge case thing, but not sure how" rather than "Help, I have broken this thing"
- Get their environments deployed faster
- If they have deployed Proxmox VE before taking the training, they are extremely likely to make significant changes to their deployments after the training
Note: To be clear, we are not saying you should take the training with Weehooey. All of the Proxmox-authorized trainers are skilled and experienced people who do a good job of training. Our team has three people who have received training from other Proxmox-authorized trainers (including me). You should take the training with one of them.
You're mentioning a lot of already made decisions, but no reason how you got to those. So just a few thoughts:
Cluster sizing depends on load, current load, expected near future load, load to cope with during failures.
Same for storage, including thoughts about latency, data integrity, redundancy, backup.
What I am missing especially in your list are disaster recovery and backup (aside of many other details).
You're asking for best practices or lessons learned: Why not make a plan and then ask an expert to have a look on it and make suggestions? This might be even more efficient, if you have attended a training before like weehooey-bh stated.
oh for sure. but I think you're concentrating on the wrong thing.
why? whats wrong with what you have now? "as possible" is a, forgive me for my bluntness, a stupid metric. if I were you, I'd start by asking the question "what are the goals, and am I meeting them?" there's no point in trying to exceed goals especially if you haven't defined them.
Again, why? what is the relevance of this number?
you seem to obsess about technology instead of application. technologies are tools- until you have something to nail you dont need a hammer.
Here's the best practices from my perspective- throw away all your ideas about technique and start, clean sheet, on what it is your after- function, minimum performance requirements, fault tolerance, etc. I think training is not a misplaced idea at all, especially since training usually follows usecase/application.