Declarative workflow breaks on YugabyteDB

[lalloni]

Declarative workflow breaks on YugabyteDB

[a8m]

Adding support for a new database/driver is a quite complex process on our end. It requires us to integrate it into our CI pipelines and commit to supporting it across all types and syntax — including all our non-OSS parts. In addition, YugabyteDB does not support other PostgreSQL features that require more changes, and I believe an implementation for such driver should be different and not change the Postgres one.

Also, our HCL cannot contain unsupported options, and Yugabyte would need to have different file extension in our editor plugins (VSCode, JetBrains), similar to what we have for MariaDB or MySQL.

Currently, this is not on our upcoming tasks, as we haven’t received any customer requests for it. However, I believe we will eventually add it, and we may reach out to the Yugabyte team before we’ll start to work on it.

Thank you for your suggestion, but I'll decline it at this stage.

[giautm]

Hey, thank you for reporting this issue. But I have checked their document for syntax. It looks like Atlas still generates the correct syntax for lsm

• https://docs.yugabyte.com/preview/api/ysql/the-sql-language/statements/ddl_create_index/#create-index
• https://docs.yugabyte.com/preview/api/ysql/the-sql-language/statements/ddl_create_index/#access-method-name

They don't mention anything special about the lsm method, and it looks like USING lsm should be valid to use.

[lalloni]

It should be valid, I agree.

Since the error message mentions it, is it possible that lib/pq could be rejecting it?

[lalloni]

However, in YugabyteDB, lsm is the default index type, so there is a possible workaround on Atlas by not including the USING clause when the DB is Yugabyte, and the inspected or specified index type is lsm.

I quickly tested that, changing Atlas in a few places, and it seems to work.

Please take a look at 8cce007

If adding such a change to Atlas makes sense, I can improve the patch as needed and post a PR here.

[a8m]

Thanks for the suggestion, @lalloni. We currently do not have official support for YugabyteDB, but it's on our roadmap.

Since this is duplicate of #2420, I think we can close this issue and track the progress there.

[lalloni]

Thanks, @a8m.

I saw #2420 before, but it had no detail, plan, or code; it was just an unanswered question. So, I moved forward to making some specific progress here.

I'm not sure how to take your last comment in the context of the patch I shared earlier.

Should we pursue the idea of getting it in as a small change that at least makes it work?

And again, I can offer time from my side to improve it in any way you believe it should be improved to get it integrated back.