Managing system performance

Outline of typical troubleshooting procedure

Checking Procedure Results Action
Error Messages Check non-graphical console screen filesystem full Move or remove files from that filesystem
Error Messages Check non-graphical console screen inode table overflow Increase number of inodes for that filesystem
Paging sar -pgruwd bad response, %(wio + idle) high, freemem often low, pfree/s and vscan/s high, swpots/s > 0 often, high disk %busy Increase memory
Disk Usage sar -udb bad response, high %(wio + idle), subset of disks have high %busy, avwait, or avque Balance disk load (move users)
Disk Usage sar -udb bad response, high %(wio + idle), all disks have high %busy, avwait, or avque Add disks
Buffer Usage sar -udb %(wio + idle) high, %rcache < 90, %wcache < 65, bread/s, bwrite/s high Increase BUFHWM
Modem Interrupts sar -y mdmin/s > 1 Repair modems, terminals, lines
Table Sizes sar -vt ov > 0 fail > 0 Increase table sizes
Job Schedules ps -ef Large batch jobs, non-essential commands or users Reschedule jobs or move jobs to other systems

 Checking           Procedure        Results           Action
 Error Messages     Check non-       filesystem full   Move or remove files
                    graphical                          from that filesystem
                    console screen
 Error Messages     Check non-       inode table       Increase number of
                    graphical        overflow          inodes for that
                    console screen                     filesystem
 Paging             sar -pgruwd      bad response,     Increase memory
                                     %(wio + idle)
                                     high, freemem
                                     often low,
                                     pfree/s and
                                     vscan/s high,
                                     swpots/s > 0
                                     often, high
                                     disk %busy
 Disk Usage         sar -udb         bad response,     Balance disk load
                                     high %(wio +      (move users)
                                     idle), subset
                                     of disks have
                                     high %busy,
                                     avwait, or
                                     avque
 Disk Usage         sar -udb         bad response,     Add disks
                                     high %(wio +
                                     idle), all
                                     disks have high
                                     %busy, avwait,
                                     or avque
 Buffer Usage       sar -udb         %(wio + idle)     Increase BUFHWM
                                     high, %rcache <
                                     90,
                                      %wcache < 65,
                                     bread/s,
                                     bwrite/s high
 Modem Interrupts   sar -y           mdmin/s > 1       Repair modems,
                                                       terminals, lines
 Table Sizes        sar -vt          ov > 0 fail > 0   Increase table sizes
 Job Schedules      ps -ef           Large batch       Reschedule jobs or
                                     jobs, non-        move jobs to other
                                     essential         systems
                                     commands or
                                     users

Next topic: Samples of performance management procedures
Previous topic: Investigating performance problems

© 1999 The Santa Cruz Operation, Inc. All rights reserved.
UnixWare 7 Release 7.1.1 - 5 November 1999