Now I have found that we've few indices which are not created by default. Question is if they are worthy for everyone, or if I should just mention them in https://docs.pagure.org/koji/database_howto/

    "task_by_method" btree (method)
    "task_by_owner" btree (owner)
    "task_by_owner_no_parent" btree (owner, parent) WHERE parent IS NULL
    "task_completion_time" btree (completion_time)
    "task_create_time" btree (create_time)

Should this be 1.22 to 1.23?

Should this be 1.22 to 1.23?

updated

Apart from the cut-and-paste error with the versions, the current change looks fine for what it covers.

As far as the other indexes, I wouldn't be surprised if some of them are effectively redundant. I'm fine with adding more, but it seems like we should do a little due diligence to verify they are actually currently helpful.

Also, I wonder what indexes other Koji instances might have added themselves as optimizations.

Apart from the cut-and-paste error with the versions, the current change looks fine for what it covers.

As far as the other indexes, I wouldn't be surprised if some of them are effectively redundant. I'm fine with adding more, but it seems like we should do a little due diligence to verify they are actually currently helpful.

Also, I wonder what indexes other Koji instances might have added themselves as optimizations.

Maybe we could provide a schema_extra.sql for the optimization and other non-essentials?

For what it is worth, depending on the size of the indexes, I'd just make them by default. Smaller sites may not see any performance boost, but they also might not see any real cost. Larger sites would automatically benefit without any need for extra work.

Commit c1735bab fixes this pull-request