I've come across some discussions suggesting that running databases in Kubernetes is a bad practice. Many folks recommend opting for managed database services like AWS RDS or ElastiCache instead. I'm curious about the actual advantages and disadvantages of running databases in Kubernetes. What's the real deal?
6 Answers
If you use a database operator like CNPG, it can give you an experience that's pretty close to using RDS. The trick is to ensure that your storage backend is low-latency—avoid NFS if you can. It can definitely work if set up properly!
Honestly, using an NFS backend is okay for small workloads, as long as everything is set up right. We’ve done it without major issues.
The real issue is that you don’t have a managed RDS option for on-prem solutions. It’s about choice—do you want managed databases with old technology or something like CNPG?
Exactly! In larger organizations, you're often stuck choosing between outdated managed services or dealing with a custom solution like CNPG.
But even with custom setups, do you really end up keeping the databases in K8s for long? That's a big question.
To be honest, for many, it's just about having their databases accessible and integrated within their K8s deployments, especially for smaller workloads where maintaining a managed database might not be justifiable.
Well said! It simplifies lifecycle management for those cases, especially if it's just a temporary solution or a cache.
Absolutely! For small-scale applications, the overhead of managed services can feel unnecessary. Keeping everything under K8s can make things more streamlined.
At the end of the day, a lot of the negativity around running databases in K8s seems to come from those trying to push managed services. Some people just prefer the flexibility K8s offers, despite the risks.
Right? Managed DB supporters often overlook that Kubernetes has matured, and many setups can actually operate effectively if done correctly.
And that’s crucial! If you know what you're doing with K8s, the downsides aren’t as daunting as some might claim.
Running a database in Kubernetes might not be a problem if you have the right know-how. Some operators can really facilitate this, but when things go south, that’s when you’ll wish you had experts on hand.
I’ve seen a lot of organizations without the necessary expertise get into real trouble when their K8s setups fail. It’s not pretty, trust me.
Exactly! While managed databases make life easier, you often still need to know what you’re doing, even with those services. It’s a common misconception that they solve *all* problems.
Sometimes this choice boils down to money—some clients just won’t pay for managed services. If they’re okay with potential downtime, they’ll stick with running their own databases.
True! Clients often weigh cost versus necessity, and sometimes they decide that managing their own infrastructure is worth the risk.
Exactly! If the risk is low and they can save cash, it’s a gamble a lot of companies are willing to take.
For our needs, using Stackgres with dedicated nodes has worked well, even if it’s not the ideal setup for production environments.