I don't believe the policy interval has changed. There are multiple subtypes though, some of which may be useful to collect. If you look in the SMF manual, you will see that there are several records that are written "every policy interval for each xxxx". In my experience, turning off at subtypes 1, 2, and 3 will avoid most of the volume, although not all of the remaining subtypes are necessarily directly useful to customers either. Of the remaining subtypes, I most often use the subtype 6s. But my approach to SMF data is to collect it all, until something becomes completely unbearable, just in case I need the data to analyze a problem. My only exclusions at the moment are: NOTYPE(92(10,11),99(1,2,3))
I just checked one of my busy systems from an hour yesterday afternoon and it generated just under 3 type 99s per second with an average length of 818 bytes. Your results will vary. It would be nice if the report showed the subtypes and the volume by data size instead of just record counts, instead of having to write your own code, or use somebody else's to determine that. Cheryl Watson wrote about some options: www.watsonwalker.com/SMFCounts.pdf Scott Chapman On Wed, 22 Aug 2012 13:06:38 -0700, Mark Yuhas <[email protected]> wrote: >constructed ones is more than I expected. I reviewed the SUMMARY >ACTIVITY REPORT and discovered an inordinate amount of Type 99 records. > >I did some calculations for each LPAR and discovered that, on average, >there are 5.16 Type 99 records being written every second. Not one >every 10 seconds. Either the policy interval time length has changed or >these Type 99 records are being written by some other component. > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
