gh-ost VS vitess

I’ve been through multiple incidents where everything worked fine in the testing environment but ended up locking the production database for minutes when deployed. A category of open-source tools called OSC (Online Schema Change) exists to mitigate such pain, like gh-ost used by GitHub and OSC used by Meta. They work by creating a set of "ghost tables" to apply the migrations, copy over old data from the original tables, and catch up with new writes simultaneously. When all old data is migrated, you can trigger a cutover to make the "ghost tables" production. Check the post below for a great introduction and comparison:

Sounds like it's basically explained in the gh-ost readme https://github.com/github/gh-ost#how

I think it amounts to "use views to decouple access to the table with a fixed interface" and "use triggers for migrating data between tables"

Gh-ost is the new hotness. Simple to use and lots of great features: https://github.com/github/gh-ost

If the performance of the db is a concern during migrations (locking, high cpu consumption for large writes) there are tools that can help and do similiar to what your describing but with the benefit that they are battle tested tools. This one spring to mind https://github.com/github/gh-ost there are other options as well and its worth reading the trade off docs

Not sure which version of MySQL you're using, but one approach would be to use a tool like pt-online-schema-change (from Percona) or g-host -- which will create a duplicate table and then swap it in place of the original table. It's a safer approach when operating in production environments. Here's a good comparison of the tools many people use https://planetscale.com/docs/learn/online-schema-change-tools-comparison

No, especially on large tables with billions of records. They make online schema changes impossible. More details: https://github.com/github/gh-ost/issues/331#issuecomment-266...

Tip #4: Consider slow-running migrations. Some tables can be so large that the traditional migration way is simply not a viable option for them. In such cases, you can consider embedding the data migration code right into your application, or use a special utility like GitHub's online schema migration for MySQL. A slow-running migration can work in production for days or even weeks. It gradually converts the data by small chunks, so you can carefully balance the load on the database while making sure that it doesn't cause slowness or downtime.

GitHub gh-ost

Actually I never tried but I was scared by the small print of GH not using RDS themselves [1] and Ghost relying on lower-level features that might be not easily available in RDS. Also I had the impression you have to setup a normal non-RDS replica attached to your RDS master?

[1] https://github.com/github/gh-ost/blob/master/doc/rds.md

With Vitess likely merging a lot of its binaries into a single unified binary: https://github.com/vitessio/vitess/issues/7471#issuecomment-...

... it would be a wild future if Vitess replaced the underlying MySQL engine with this as long as the performance is good enough.

Thank you for the compliment!

We recently started adding support for CTEs in Vitess! You can check out https://github.com/vitessio/vitess/pull/14321 if you want to see some technical details of the implementation.

For now, we have added preliminary support by converting them to derived tables internally, but we believe that we need to make CTEs first-class citizens themselves of query planning. Once we make that change, we can look towards supporting recursive CTEs.

This however will take some time, but then, all good things do!

Why would it be a Google project? https://github.com/vitessio/vitess

This is great news. I strolled around https://github.com/vitessio/vitess/issues/12967.

Are there any public discussions of more trade-offs vitess has to make to enable fks?

MySQL + Vitess I noticed Slack gets by using MySQL because they're using Vites. From Slack's post (https://slack.engineering/scaling-datastores-at-slack-with-vitess/) it seems like they choose Vites because it facilitated a smooth transition because it's built on top of MySQL.

This sounds very similar to the connection pooling done by vitess https://vitess.io/.

https://github.com/vitessio/vitess/issues/12967

A lot of this is possible thanks to the magic of Vitess.

"Vitess is a database clustering system for horizontal scaling of MySQL."

https://github.com/vitessio/vitess