Review impact on customers who are using LoadMode.DBT_LS cache

[tatiana]

We've worked closed with customers to implement #1014, and over two weeks, we confirmed their AIrflow metadata database was working well and without any significant roles, after the introduction of the LoadMode.DBT_LS cache.

This ticket is for us to remind us to review their database performance after a month this feature was rolled out.

[tatiana]

Shared with relevant parties in:
https://astronomer.slack.com/archives/C03SJKV2BK5/p1723641759349269?thread_ts=1717618150.697769&cid=C03SJKV2BK5

I just checked this to see how the database is coping with the solution; after a few weeks, it was deployed live.

The CPU utilization of the Airflow db has increased by ~15% since adopting the dbt ls cache, in early July.

The memory utilization does not have a significant difference.

The disk utilization seems unaltered.

There has been an increase in the Network I/O.

We're implementing an alternative to the Variable-based caching, that should help with the metrics that were changed in the Airflow db:
#1147