Hacker News new | ask | show | jobs
by siscia 3146 days ago
Since we are talking about postgres, please let me go a little OT

Is there any interests in a RDS as a Service? So basically setting up and running a completely fault aware postgres cluster in any infrastructure, either public or private?

8 comments

Oh dear sweet lords in the many heavens yes. I have literally turned down devops transition and general system administration work before because they refused to use RDS or equivalent service like Compose.io due to internal corporate policy. Life is too short for me to ever babysit another database server through the painful process of a carefully orchestrated rolling capacity upgrade by way of deliberate use of replicas and failover. (Apparently this has gotten better in newer Postgres versions, but it's not been my job to keep on top of the implications of these things for several years now)
Pretty much the same pain point that I experience.

Thanks for your feedback :)

Pivotal & IBM maintain a BOSH release of PostgreSQL for this purpose: https://github.com/cloudfoundry/postgres-release

Since it uses BOSH, you can deploy to a wide range of targets. OpenStack, vSphere, AWS, Azure, GCP and I forget what else.

Disclosure: I work for Pivotal.

A lot of interest. Setting this up is too much sysops for someone who wants to spend his time developing code.
Thanks for your feedback :)
What seems to be missing in the market is the ability to only pay for the setup and maintenance operations, but use my own existing cloud account and resources instead of running multi-tenant or in some other company's cloud account.

Basically as if I hired a contractor to install, monitor and upgrade, but automated. Existing services charge too much since they resell VMs and storage, while also being less flexible with access and performance.

There's also the rise of Kubernetes (with operators, helm charts and persistent storage) that takes away much of the complexity. By version 2.0, it should be able to easily make any legacy single-node system into a fault-tolerant service.

Exactly!

Ideally, I would just need an SSH key inside your machines and the capabilities to open an ssh tunnel inside the firewall to scrape metrics.

Ideally, the metric should get exposed back to the customer.

I am not a big fan of containers when working with data that are irreplaceable. But the use k8s may really help.

We would be interested, along with other datastores like ElasticSearch, Redis, etc.

AWS, Azure and Google all have various resource organization systems now so that the environment can be isolated but still within our overall account. We run on Google and it would be nice to have a separate project for managed databases while taking advantage of our existing billing arrangements and private network.

I will make sure to keep you updated :-)
Yes, but the trick with it is pricing it so that I have a reason to use it instead of just using RDS.

If I could get something like that on Digital Ocean I’d be all over it.

Well, Aiven seems to support DO: https://aiven.io/postgresql
Out of interest, could you support MySQL? Or more specifically MemSQL?
MySQL and MemSQL are very different, are you looking for a data warehouse specifically?

They are closed-source and require enterprise licensing based on RAM quota so it's not simple to do automated cloud provisioning. They do have their own MemSQL cloud offering so you might inquire into that. Also MemSQL Ops is probably the easiest and most reliable operations software for any database, it just takes a few clicks to install and upgrade your cluster.

This is really just an idea to solve a pain point of mine that I guess is shared among us...

By the way I don't have any experience with MemSQL

I imagine there is interest in it since Aiven provide such a thing.
Aiven works only with public clouds...
Yep!