One last thought, for true 'best humanly possible' service uptime, you really should be looking at high availability (HA) clustering. That's how serious high availability is done in industry, multiple nodes with as much redundant duplication as possible. When properly implemented, and carefully managed, you can take cluster nodes in and out of production with no more than about a second or two of disruption to the services running on them. That lets you do necessary admin tasks on the cluster, including reboots for things like kernel updates, with minimal impact to services. It also covers you for individual node failure (either hardware or OS failure). In normal production, the cluster nodes can share the service workload between them, so it's also good for scaling the services.
Statistics: Posted by Murph9000 — Fri Nov 08, 2024 7:48 am