Comment Re: Better go kick WSUS into a sync... (Score 1) 178
I help develop and operate a service that makes a hefty sum by doing all those things you deride, implementation-wise. It all works quite well - well enough that if routing patching causes any customer-visible disruption, you're in for extensive analysis, paperwork, and perhaps ritual abasement before an angry VP.
Yes, yes, there are many technical problems involved with consuming "eventual consistency". In the 20th century these problems were seen as blocking, and anyway just buy a bigger DB server. But the 20th century was along time ago, and while there's still a need for a transactional store, most problems can be solved without one, given sufficient thought - and at sufficient scale, it's really worth figuring out how.
Not that safe patching is incompatible with SQL, of course. In my last job we routinely pushed patches to farms of many thousands of SQL servers, and again if there was any disruption visible to the mid-tier, important people would become seriously angry about that, and we didn't use fancy servers, beyond RAID controllers (and even that concession I abhorred). It's always safe for a single server to fail, or be rebooted for maintenance, and if two servers holding your primary copies of the same data should fail, you better have taken serious, well-reviewed steps in planning to limit the number of DBs affected and the minutes of data lost and the minutes until you're back up.
And even that, which was a nice system, feels outdated now that Amazon went and announced this, which productizes the modern SQL DB and wraps it up in a pretty bow.