HP 3000 Manuals HP 3000 Manuals
Log File Resizing Strategies When raw log files are filled with data, they are rolled back automatically by the SCOPE(XL) data collection program. When a log file is rolled back, a new log file the same size as the old log file is built, the latest 75 percent of the data from the old file is copied to the new file, the old file is purged, then the new file is renamed with the old file's name. The net effect is to discard the oldest 25 percent of the data to make room for new data. Case 1 The simplest strategy for managing raw log files is to do nothing and let the automatic process take care of the logs. You might want to select the amount of data each log file can hold to be sure that you have a specific amount of raw data available at all times. After logging data for several days, you can use the UTILITY program's RESIZE command to size each raw log file. Choose a size that will provide the desired amount of data when the log file is 75 percent full without using excessive disc space. For example, if you want to have 30 days of global data available at all times to allow you to prepare one-month global graphs, size the global log file to hold 40 days of data (75 percent of 40 equals 30).
Hint Especially on MPE V host systems, avoid sizing raw log files too large, since this can slow access to the data in those files. MPE/iX host systems are not as sensitive to the size of raw log files.
Case 2 Since rolling back a log file can use significant system resources, you may want to schedule when it will occur. If you fill a log file during the prime-time shift, when maximum system usage occurs, the roll-back operation might compete with other processes for existing resources. This can cause online response times to degrade for as long as 5 minutes on heavily-loaded systems with large log files. Also, you may miss logging important performance data because data collection is suspended during the log file roll back operation. You can instruct the SCOPE(XL) collector to resize log files at scheduled times by adding the MAINTTIME directive to the PARM file. The SCOPE(XL) program will resize a log file at the specified time if the program expects the file to become full within the next 24 hours. Since log records may accumulate at different rates depending on your system's activity, scheduled maintenance may resize log files too early or too late. In such cases, you might want to control log-file resizing more closely. You can avoid unscheduled log file roll backs by using the UTILITY program to ensure that the raw log files contain sufficient empty space. Use the UTILITY RESIZE command to perform a scheduled log-file roll back. For example, once a week you can run a batch job that ensures there is enough room in the raw log files to hold another week of data. If there is not enough empty space, the log file will be rolled back, and old data will be discarded to provide the desired space. For example, :JOB MAKEROOM,SCOPE.SYS,SCOPE :RUN UTILITY SCOPE KILL (To turn the data collection program off.) LOG LOGGLOB SCAN (To determine how much data is logged daily.) RESIZE GLOBAL EMPTY=10 MAYBE RESIZE APPLICATION EMPTY=10 MAYBE RESIZE PROCESS EMPTY=10 MAYBE RESIZE DISCSPACE EMPTY=10 MAYBE EXIT :STREAM SCOPEJOB (To restart the data collection program.) :EOJ The MAYBE parameter is added to the RESIZE command to force resizing only if the empty space currently in the log file is less than the requested amount. Ten days is chosen instead of seven days in case the amount of data logged each day increases slightly. Because this operation might involve discarding data from raw log files to make room for new data, you should archive existing data before proceeding with this action.