Chapter 13. Proxmox Production-Level Setup

So far in this book, we have seen the internal workings of Proxmox. We now know how to properly set up a fully functional Proxmox cluster. We discussed Ceph—a robust and redundant shared storage system—and how we can connect it with Proxmox. We also saw what a virtual network is and how it works with the Proxmox cluster.

In this chapter, we are going to see which components play a crucial part in making a Proxmox cluster production-ready with multi-layer redundancy, good performance, and stability. We are going to cover the following topics:

• The definition of production level

• Key components of a production-level setup

• Entry-level and advanced-level hardware requirements

Throughout this chapter, you will notice that we have used user-built hardware configurations instead of readymade brand servers. The purpose of this is to show you what sort of node configuration is possible using off-the-shelf commodity hardware to cut cost while setting up a stable...

Defining production level

Production level is a scenario where a company's cluster environment is fully functional and actively serving its users or clients on a regular basis. It is no longer considered as a platform to learn Proxmox or a test platform to test different things on. A production-level setup requires much advanced planning and preparation because once the setup is complete and cluster has been brought online, it just cannot be taken offline completely at moment's notice when users are dependent on it. A properly planned production-level setup can save hours or days of headache. If you are still learning Proxmox, you might want to set aside hardware to practice on so that you can hone your skillset before attempting a production-level setup. In this section, we are going to cover some of the key components or characteristics of a production-level environment.

Key components

The following key components should be kept in mind while planning for a production-level cluster setup...

Sizing CPU and memory

A question often asked when it comes to creating virtual environment is how much CPU or memory will be needed in each node and how much to allocate per virtual machine. This is one of those questions that are very open ended because their answer varies greatly from environment to environment. However, there are a few pointers that need to be kept in mind to avoid over-allocation or under-allocation.

It is a fact that we will and often do run out of memory much sooner than CPU for a given Proxmox or any other host node. From the usage of each VM on the Proxmox nodes, we can determine the RAM and CPU requirement on that node. In this section, we are going to go over the factors that will help us to decide on CPU and memory needs.

Single socket versus multi-socket

A multi-socket node will always have better performance than a single socket regardless of the number of cores per CPU. They work much efficiently distributing VM work load. This is true for both Intel and AMD architectures...

Production Ceph cluster

As mentioned throughout this book, Ceph is a very resilient distributed storage system that pairs well with Proxmox to store virtual disk images. There are a few key factors that make Ceph cluster a good choice for production-level virtual environment.

Forget about hardware RAID

When it comes to Ceph node and cluster, we can forget about hardware-based RAID. Instead, we have to think multi-node or clustered RAID. That is because of the distributed nature of Ceph and how data is dispersed in all the drives in the cluster regardless in which node the drive is. With Ceph, we no longer need to worry about a device failure in a particular node. Ceph performs best when it is given access to each drive directly without any RAID in the middle. If we are to place drives in RAID per node, we will actually hurt Ceph immensely and take away everything that makes Ceph great. We can, however, still use the RAID interface card to implement JBOD configuration or to be able to connect...

Liquid cooling

In this solution, computer equipment is cooled using liquid for heat transfer instead of air. Liquid is 1000 times better at heat transfer than air. We can effectively remove heat directly from IT equipment and transfer it with great ease out of the facility. Liquid cooling takes away the hassle of running large HVAC system, thus saving enormous costs and reducing noise significantly. Liquid cooling requires no internal fans, thus we can increase server density per rack 10 fold. Liquid cooling is the future as more and more IT facilities are realizing its full potential. By using liquid cooling, we can also decrease our energy consumption—reducing carbon footprint enormously. There are different liquid cooling solutions available on the market.

Total immersion in oil

IT equipment is totally submerged in mineral oil. Hot oil is pumped through liquid-to-liquid heat exchanger, where the heat is carried away using water to an outside cooling tower. Water and oil never have full...

Real-world Proxmox scenarios

Equipped with all the knowledge we have gathered from the previous chapters in this book, we are now ready to put all the pieces together to form complex virtual environment for just about any scenario that we are going to be called for. Some set of scenarios are given in the next section to build networks using Proxmox for various industries. At the end of the chapter, you will find network diagrams of each scenario given in first part of the chapter.

Some scenarios have been taken from real-life production environments, while some are theoretical to show how complex networks are possible with Proxmox. You can take these network models and use them as they are or modify them to make them even better.

We hope that through these network scenarios and models, you will start seeing Proxmox from a whole new point of view and be fully prepared to face any level of virtual infrastructure you are challenged with.

Tip

While analyzing these scenarios, keep in mind that the...

Summary

Virtual environments are very flexible, so there is no one-network-fits-all configuration. Each network will be unique. The components and requirements described in this chapter are mere guidelines for how to take the correct approach to plan for a production-level Proxmox setup. We saw some of the requirements of a production-level setup, and we covered how to allocate CPU and memory resources properly for both Proxmox host node and virtual machine itself. We also discussed how to give Ceph storage the best chance of providing redundancy along with performance. Finally, we saw how to cool equipment efficiently by leveraging liquid cooling, thus increasing Proxmox computing node density per rack while saving energy.

We also saw some real-world scenarios of Proxmox in action in different industries. We hope this will aid you in your quest to find that perfect balance between performance and budget that all network administrators crave.

In the next chapter, we are going to see some real...