Self-Hosted Engine Red Hat Virtualization Architecture

Latest response

Not sure if I'm understanding Self-Hosted Engine Red Hat Virtualization Architecture.

PREFACE - Hosts that can run the Manager virtual machine are referred to as self-hosted engine nodes. At least two self-hosted engine nodes are required to support the high availability feature.

Question, does this mean I can run RHV-M without having a second server, but in order to use the "high availability feature" I do need a second server?


What a moment, I think I understand this based on this reading Table 1. Red Hat Virtualization Key Components,

Red Hat Virtualization Manager: The Manager is installed on a physical or virtual machine running Red Hat Enterprise Linux. Hosts: Red Hat Enterprise Linux hosts (RHEL hosts) and Red Hat Virtualization Hosts (image-based hypervisors) are the two supported types of host.

So were basically talking about two VM's as the self-hosted engine nodes and therefor I only need 1 physical server if I want to use the "high availability feature" correct?

So the hosts for using the

Also, the initial base installation of RHEL is the default Server + GUI, so I assuming not to make any other selections and just install the default correct?


Figure 1.1 in the Product Guide probably explains it best. The Manager is just a VM that runs like all your other VM's on the host (it has a few special features though). If it's running on Host1, and that Host dies, it can easily be restarted on Host2. If you only have one host, it can't be restarted anywhere else, so the solution is not "highly available". But then again, none of your other VM's have anywhere to run either.

Host can either be regular, or 'self-hosted engine' (SHE) hosts. Regular hosts can only run regular VM's. SHE hosts can run the special HostedEngine VM, as well as regular VM's

The best number of SHE hosts is 2-4. So for small environments, (2-4 hosts), make all of them SHE. If larger environments (like 20 hosts), just stick to 4 SHE hosts, and spread them sensibly for resilience (hosts on different power boards, network switches, etc) and make the rest regular hosts.

SHE hosts need to be in constant communication with each other to make sure only one instance of the Manager is running. So having too many SHE hosts creates a larger communication overhead, and over-complicates the SHE cluster.

If you are using RHEL hosts as hypervisors, Install Docs mention this should be done as a 'basic install', which does not include the GUI (it's not needed on RHV hosts)