@mcr314 oh it absolutely doesn't, but it would be mighty convenient if it all was on a single box

@penguin42 Yeah the 8S machines are also a considerably more magic, See https://community.hpe.com/t5/servers-systems-the-right/defying-the-law-of-diminishing-returns-hpe-superdome-flex/ba-p/7012970

I suspect this system is utterly haunted, also at some point you start to have so many DIMMs etc that you are basically asking for problems (yes, I know there is RAS etc, but I don't think Intel's version is bomb proof enough in the same way other non x86 systems managed to get it)

@penguin42 yeah the problem with 8S though is the performance sucks if you ever make the mistake of intra NUMA memory access, and not just "oh that kinda hurts" in the way it is on 2S systems... It really sucks

@jeroen I've had the complete opposite in my career so far, one big-ass-box can run laps around the availability/integrity of clusters of machines, it's a question if you would rather take one big "bang" outage every ~5 years or tiny outages every couple of months

@mcr314 @jeroen I guess I'm generally a bit absolutist (when I can be) in this regarding that I don't actually see active/standby as a good option and I will try as hard as I can to avoid it, active/active or you may as well go restoring a backup, since the number of times I've seen such standby operations go wrong or cause extra work/surprises etc, or just not work correctly when the standby needs to become active

@FritzAdalis @mcr314 @jeroen Yeah but that is a huge commitment right? I understand that not everything can be active/active but _surely_ most systems can be these days, or people can accept downtime for rare "shit's fucked" situations.

Though if it's something like a EHR system, then yeah I can see how this might be a more complex take...