When zabbix_server works with the DebugLevel=3 it reports to the log this message:
"Deleted <NNN> records from history and trends"

But here are not included values deleted by housekeeper for the items already deleted from configuration (table "housekeeper").
1. This lack of information could confuse when troubleshooting of DB performance.
2. It's not consistent that the housekeeper is reporting count of the deleted outdated values but does not report count of values for items already deleted from configuration.

Try to imagine a situation when some user has performed "Unlink and clear" action with some big template linked to the several hosts and in the nearest hour he felt some problem with the DB performance and he thinks - what happened ?

Here is part of DebugLevel=4 (I added several EOL for better view):

15500:20111031:150250.029 End of housekeeping_history_and_trends():0

15500:20111031:150250.029 Deleted 0 records from history and trends

15500:20111031:150250.029 In housekeeping_process_log()
15500:20111031:150250.029 query [txnlev:0] [select housekeeperid,tablename,field,value from housekeeper order by table name]
15500:20111031:150250.029 query without transaction detected

15500:20111031:150250.029 query [txnlev:0] [delete from history where itemid=22578 limit 100]
15500:20111031:150250.044 deleted 100 records from table 'history'

15500:20111031:150250.044 query without transaction detected

15500:20111031:150250.044 query [txnlev:0] [delete from history_uint where itemid=22577 limit 100]
15500:20111031:150250.060 deleted 100 records from table 'history_uint'

15500:20111031:150250.060 End of housekeeping_process_log():SUCCEED
15500:20111031:150250.060 In housekeeping_events() now:1320062561
15500:20111031:150250.060 query [txnlev:0] [select event_history from config]
15500:20111031:150250.060 query [txnlev:0] [select eventid from events where clock<1288526561]
15500:20111031:150250.060 End of housekeeping_events():SUCCEED

15500:20111031:150250.060 In housekeeping_alerts() now:1320062561
15500:20111031:150250.060 query [txnlev:0] [select alert_history from config]
15500:20111031:150250.060 query without transaction detected
15500:20111031:150250.060 query [txnlev:0] [delete from alerts where clock<1288526561]
15500:20111031:150250.060 deleted 0 records from table 'alerts'
15500:20111031:150250.060 End of housekeeping_alerts():SUCCEED

15500:20111031:150250.060 In housekeeping_sessions() now:1320062561
15500:20111031:150250.060 query without transaction detected
15500:20111031:150250.060 query [txnlev:0] [delete from sessions where lastaccess<1288526561]
15500:20111031:150250.061 deleted 0 records from table 'sessions'
15500:20111031:150250.061 End of housekeeping_sessions():SUCCEED
15500:20111031:150250.061 sleeping for 3600 seconds

(in this example "MaxHousekeeperDelete=100")

As you see in the function "housekeeping_process_log()", where the values are deleted from the tables history, history_uint (and not only), no messages are added for the LOG_LEVEL_WARNING.

I ask to add these messages to log. Maybe individually per each table, maybe summarize history+trends, I don't know how is better.

If any such values were not deleted then needn't to report any messages.