Customer success sometimes runs manual operations on customer databases, such as CLUSTER in the context of PDB-959. These changes can be lost during upgrades, so to ensure the effects are retained we should formalize a way to track them and hook them into the migration process.
Charlie SharpsteenNick Walker Care to comment on this? I could see adding a "post-migrate.d" dir full of sql scripts or something, but would like to collect some actual use cases first.
The one thing I can think of off the top of my head is that an upgrade might be a good window to run some general maintenance like re-building indices. Nick has some of this in his pe_databases module:
The original case that's mentioned around CLUSTER turned out to not be specific to CLUSTER but to rebuilding the table. I'm not sure we have a need for this?
We do have a need to make indexes sometimes but I think that's also better handled by puppet code.