asktom的含义maxloghistory

1) well -- the documentation seems clear on this to me: <quote> Note: This parameter is useful only if you are using Oracle with the Parallel Server option in parallel mode, and archivelog mode enabled. </quote> so, it would be saying that is has not significance outside of OPS/RAC. It controls the number of log history records available on v$log_history. 2) well, it depends. RMAN usually uses a recovery catalog, not the control file. If you are using RMAN with just the control file and no recovery catalog, it'll totally depend on how fast you cycle through and reuse reusable records in the control file. In any case -- you can always recover manually regardless (you do not lose the ability to recover, we just might not be able to supply you the filenames) 3 & 4) they shouldn't be the same. loghist (log_history is the prefered view) is controled by the maxloghistory. v$archived_log is controlled by the control_file_record_keep_time -- those records are among the circularly reusable records. We'll dynamically expand the control file to hold 7 days worth of archive log information for you (if you have it set to 7). On the 8th day, we just overwrite the old records and reuse their slots. So, one is retained by a physical count (maxloghistory) and shows the history of the logs. The other is retained by a number of days -- keep the history for the last N days -- and will cause the control file to grow as needed. As for why you are seeing really old records in v$archived log -- you must have had a 7 day period in which you generated LOTS of archives, more then you did in the recent past. That caused this section to grow larger to hold it. We won't shrink it back -- hence you have more space then you need to hold your current 7 days worth of information. The 7 is a MINIMUM retention time. It does not mean we purge records older then 7 days. 5) set your control_file_record_keep_time to be at least one day GREATER then the period of time between your backups, else there is a chance that an archive record gets aged out during the backup which thoroughly confuses the situation.