QSF Monitor

From PTAGISWiki

The QSF is the Ingres Query Storage Facility which interacts with the Parser Facility. Sometimes it runs out of memory.

The QSF Monitor keeps us apprised of the usage patterns of QSF.

The operation of the monitor is overly complex as a result of the difficulties in getting proper permissions for scripted tasks that require elevated permissions. Here is the breakdown of how it currently works.

Flow of QSF Monitor control

• cron job for user rday runs on snapper that fires a perl script (qsf.pl)
• a network mounted file system on blueback is examined to pick up the latest usage value
• another perl script on snapper takes that value and inserts it into an RRD file on snapper. It then copies the file to pitblade for web display.

On blueback

• a cron job for user rday runs a shell script (qsf.sh) which sets environment variables and launches an expect script (qsf.exp) that returns the memory usage of qsf and places it in a text file

Managing QSF usage

Doug tried using a tool to mark the QSF as reclaimable, but it didn't recover as much space as we would like.

Hi, Ryan -
 
I just ran the cache cleanup tool (sets trace point QS506) 
on 'blueback'. It reduced QSF cache utilization from 57% to 48%. 
Not the major improvement we'd hoped for, but still an interesting result.
 
I'll give some further thought to the problem and let you know 
when I come up with something else to try.
 
-Doug-

Removing idle sessions

• log in to blueback as ingres
• cd to /home/ptagdev/ptagis3/dba_chores/ServerManagement
• execute ./DBA_removeIdleSessions.pl