I'm looking how to make my domain-based architecture highly available.
My understanding is that when a Host Controller is starting it's trying to contact a Domain Controller. When the Domain Controller is unavailable, the Host Controller fails to start. I can prevent this by passing --cached-dc to the Host Controller, but this means the HC will not even attempt to contact the DC and will always use a cached version domain.xml, which effectively downgrades it to a standalone node and having --cached-dc in start script is not an option.
This means the DC is now a SPoF and I should make it as Highly Available as possible. Are there any recommendations how to achive this? I'm considering having 2 DC nodes and virtual IP floating between them. I understand I need to make sure that exactly 1 instance of DC is always up and running - hopefully RHEL clustering can help a bit here. I also need to make sure the domain.xml is on a shared filesystem.
Does anyone have a better idea? Thanks a lot for any hint!