Skippr Docs

Appearance

What a Good First MySQL Integration Looks Like ​

July 2026

The smallest production MySQL setup is one connection string, one database, and a deliberate choice between selected tables and discovery.

Short Answer ​

A good first MySQL integration is one env-backed connection string, one read user, and a conscious decision about table scope. The docs support a connection_string and an optional tables allowlist, which means the smallest good production version can be either a narrow named set like customers and orders or controlled discovery when the readable schema is already trusted.

For most teams, the better first production boundary is explicit table selection. It makes the contract reviewable and keeps the first run small. If the long-term plan includes binlog CDC, that should influence how the MySQL user is created, but it should not force the first integration to do more than a clean batch read from the tables that matter.

Why Teams Struggle with This ​

MySQL integrations go wide very quickly because discovery feels convenient. In practice, a production-shaped start usually needs a smaller boundary than the whole readable database.

  • The runner must reach the MySQL host and port before table selection matters.
  • A DSN that works for the application is often too broad for analytics extraction.
  • Auto-discovery is useful, but it can pull in more tables than the first downstream model is ready to own.
  • Binlog CDC has extra requirements, so it is better treated as a second milestone after the base read is healthy.

How Skippr Handles It ​

Skippr keeps MySQL simple enough to stage safely. One MYSQL_CONNECTION_STRING unlocks the server, and an optional tables list makes the first production boundary explicit when discovery would be too wide.

That separation also helps operations. The team can rotate the credential in the DSN without changing the source contract, and it can tighten the table scope without changing the auth path.

  • Uses one env-backed MYSQL_CONNECTION_STRING.
  • Supports a named tables allowlist or full discovery.
  • Fits a dedicated read user for batch extraction.
  • Can later move to binlog CDC when the source and destination are ready for that contract.

What the First Useful Version Looks Like ​

The first useful version is one MySQL database, one read user, and one or two named tables whose shape is already important downstream. That is enough to prove the DSN, table access, and extracted columns without making the first run noisy.

If discovery is genuinely the right long-term shape, use it after the first narrow handoff already works. A production start should reduce surprises, not maximize them.