Clustered Databases in Valkey 9 blog post

[stockholmux]

Description

• Adds a blog post about numbered databases in cluster mode and a few use cases.
• Fixes bullet points in blog posts being rendered as normal text.

Issues Resolved

#334

Check List

• Commits are signed per the DCO using --signoff

By submitting this pull request, I confirm that my contribution is made under the terms of the BSD-3-Clause License.

[madolson]

I don't know how I feel about the intro. I see it has a thesis that in the past people said don't use databases, but you can use them now. It might be more straight forward to make the case that they weren't well supported and we're working to make them easier to support?

The rest of it makes enough sense. I think telling a story about how we're working to continue improve it might also work a bit better. There is an open PR for database ACLs.

[ranshid]

I think this is a very important blog. The way I see it is that this is the fist public statement of divergent philosophy we are taken from the old fork.
For this to be motivating I would like to suggest a slightly different structure that might also work:

1. I would start (as you did) with some mention of the old philosophy and how we see it today in Valkey.
2. Then, I would provide good usecases were applications can use multi-databases. things like data separation, sandboxing, rollback etc...
3. The next part (IMO) should be focused on how we extended the ability to scale with databases
4. I would now add a section about future plans, and maybe also include our strategy (which does not target full multi-tenancy support, but aims for better observability and control over "namespaces")

[stockholmux]

I think I resolved the feedback. Give it another read @madolson / @ranshid !

[asafpamzn]

Some client do not allow to update the DB post client creation. That is, the DB is immutable. We recommend in such case to use a DB per client.

It is complex in async clients or clients that supports pipelines to support updates of the DB using select command post the client creation as it may cause data corruption in edge cases. I can elaborate more

[stockholmux]

Line 88 covers most of this. I'm well aware of the issues of clients not fully managing SELECT the correct way.

Do you have a list of the clients don't support SELECT (e.g. only at connection)? I'm glad to add this to 88.

[madolson]

One material comment, I could be convinced to leave it though, but other than that looks good to me.

The blog still feels very much like you wrote it but comes across a lot more positive now, I love most of the changes.

[ranshid]

Overall looks very good!

I added some suggestions to consider