Document the list of known ignored statements in PGDUMP

[lnhsingh]

Lauren Singh commented:

Follow up to #9494, @mwang1026 commented:

Separately, and maybe it's a separate project, I do feel like we need to have a list of "known ignored statements in PGDUMP and what to do about them". Not knowing this causes pain for customers who don't know beforehand.

For example, if there is a column type that we don't support (e.g. tsvector) the CREATE TABLE statement will fail and what will be logged in the ignored statement log is tsvector: could not be parsed

We also know for example that we ignore all SET statements
Anything relating to grants and roles and permissions, because we don't want to accidentally grant permissions to a user in the destination cluster who happens to have the same username as someone from the previous cluster

As a part of this project, I want us to at least think through the structuring of such documents.

Jira Issue: DOC-1076

[ericharmeling]

As of 21.1, we have a cluster setting that allows you to turn off ignoring unsupported statements (doc'd here: #10186), so I'm not sure this is a known limitation any more?

We might need to update the PostgreSQL Compatibility page still, so I won't close this.

[mwang1026]

Which known limitation are you referring to? I think this ticket is more about the fact that we do have a definitive list of things we ignore and it might benefit customers to list them out and how to work around those limitations.

[ericharmeling]

@mwang1026

I was really just referring to lifting the "T-known-limitation" tag - it's what we use to track the known limitations that need documenting on the Known Limitations page ahead of each GA release.

I think this ticket is more about the fact that we do have a definitive list of things we ignore and it might benefit customers to list them out and how to work around those limitations.

That sounds like a great idea! I think it might be best to house that information on the PostgreSQL Compatibility page rather than the Known Limitations page. The PostgreSQL Compatibility page already has a list of unsupported/differing features.

[mwang1026]

Ah sure that makes total sense (didn't notice you had removed the tag :D)