Re:How we migrated onto K8s in less than 12 months
by fragmede on Thursday August 08, 2024 @09:44PM
(#41198126)
Why would wanting redundancy be a ZIRP? Is blaming everything on ZIRP like Mercury was in retrograde but for economics dorks?
[ Parent
| on HN ]
Re:How we migrated onto K8s in less than 12 months
by felixgallo on Friday August 09, 2024 @04:33AM
(#41199939)
Because the company overhired to the point where people were sitting around dreaming up useless features just to justify their workday.
Re:How we migrated onto K8s in less than 12 months
by consteval on Friday August 09, 2024 @11:16AM
(#41202679)
It depends on the cost of complexity you're adding. Adding another database or whatever is really not that complex so yeah sure, go for it.
But a lot of companies are building distributed systems purely because they want this ultra-low downtime. Distributed systems are HARD. You get an entire set of problems you don't get otherwise, and the complexity explodes.
Often, in my opinion, this is not justified. Saving a few minutes of downtime in exchange for making your application orders of magnitude more complex is just not worth it.
Distributed systems solve distributed problems. They're overkill if you just want better uptime or crisis recovery. You can do that with a monolith and a database and get 99.99% of the way there. That's good enough.
Re:How we migrated onto K8s in less than 12 months
by addaon on Friday August 09, 2024 @02:01PM
(#41204031)
Redundancy, like most engineering choices, is a cost/benefit tradeoff. If the costs are distorted, the result of the tradeoff study will be distorted from the decisions that would be made in "more normal" times.