Document running ANALYSE through `gitlab-psql`
Customers with large databases encountered problems with high CPU usage after upgrading to PG11. This was caused by the
analyze_new_cluster.sh script failing to run but not giving the appropriate error code. !4227 (merged) bypasses this script and calls
vacuumdb directly. With this change, the user is at least aware if the VACUUM and ANALYZE fail.
A customer reported still running into statement timeout issues when running
vacuumdb directly. e.g. #5456 (closed)
One solution to avoid the timeouts might be to run ANALYSE through
gitlab-psql with the statement timeout set to zero. (Related discussion here). We should test this and add it in a troubleshooting section of the docs.