Description of problem:

On a system with many applicable errata, the following sort can take upwards of 90 seconds:

Sort (cost=218,484.99..219,063.42 rows=231,373 width=879) (actual time=133,527.641..133,528.216 rows=21 loops=1)

    Sort Key: katello_errata.updated, katello_errata.id, katello_errata.uuid, katello_errata.errata_id, katello_errata.created_at, katello_errata.updated_at, katello_errata.issued, katello_errata.errata_type, katello_errata.severity, katello_errata.title, katel
    Sort Method: external merge Disk: 271928kB

This causes the dashboard to load slowly.

Version-Release number of selected component (if applicable): 6.2.1

Steps to Reproduce:
1. load postgres db with lots of applicable errata
2. view dashboard page

Actual results:

extremely long load time due to postgres having to sort on-disk instead of in memory.

Expected results:

page should load quickly

Additional info: Each row of the query is about 1K in size. This requires quite a bit of memory to sort when there are 200K+ rows. Note that it is not feasible to set work_mem high enough to handle this type of sort, since work_mem is used per query and even per subquery. The likely fix is to strip the errata text in the query and then select it back in if we still need it after obtaining the top 6 errata.

Secondary note: there were some other bugs in the past related to this query. This particular issue is for the same query, but the issue is different.