TOCPREVNEXTINDEX

Lund Performance Solutions

SOS Host-specific Configuration Files

SOS advice File

In the SYSTEM PERFORMANCE ADVICE portion of the Global Summary screen, advice messages are displayed based upon system activity that occurred during the current interval. The advice messages and display criteria are maintained in the SOS advice file (a portion of which is shown below) located in the /etc/opt/lps/cfg directory.
User Notification Command
 
################################################################################
—
echo >/dev/console
Comments
—
#lpstrap
 
 
#uncomment the above line to start sending snmp trap messages with
 
 
#notify information. You must configure lpstrap for you environment also.
 
 
#See lpstrap for more info.
 
 
 
Default Advice Specification Block
—
<CI01>The CPU was used a total of %s of its capacity during this interval
 
ALWAYS
 
CPU-BUSY%
Figure 8.1 SOS advice configuration file (example)

SOS advice File Configuration

The SOS program can display a single-line message for each item-name variable (a data item selected from the /opt/lps/lib/itemlist file) placed in the advice file. For a list of the data items in the itemlist file, see Appendix C .

During each current interval, SOS compares the value of each variable being monitored to the threshold criteria placed in the advice file. If the monitored value meets its threshold criteria, the message associated with that variable is displayed in the SYSTEM PERFORMANCE ADVICE portion of the Global Summary screen.

NOTE Please note that the lower and upper bounds of the thresholds for the moderate, HEAVY, and EXCESSIVE categories of each default advice message in the advice file are suggested values. It may be appropriate to adjust these values to reflect your system’s performance criteria.

Advice Message Specification Blocks

Advice message specification blocks are constructed in accordance with specific configuration rules and syntax. The rules for configuring advice message specification blocks within the advice file are listed in "Configuration Rules". The syntax of the specification blocks is outlined below using the default ME01 advice message as an example.
Example
<ME01>Page out rate reveals %s %s memory load
VM-PAGE-OUT-RATE (10-50)
VM-PAGE-OUT-RATE | 20 an | 15 a | 10 a |
VM-PAGE-OUT-RATE | 20 EXCESSIVE | 15 HEAVY | 10 moderate |
Syntax
<message-id><message-text>
item-name (min-max)
item-name [|<value1><string1>|<value2><string2>|<value3><string3>|]
item-name [|<value1><string1>|<value2><string2>|<value3><string3>|]
Where:
  • <message-id> is a unique, four-character message identification code.
  • <message-text> is the actual advice message text.
  • item-name is the itemlist value to be used to determine the text string.
  • (min-max) is the minimum and maximum item threshold values required for the message to display.
  • The last two lines in the example are each single-line text qualifiers that correspond to the text place-holder(s) (%s) in the message-text.
  • The first place-holder in the message-text corresponds to the first text qualifier in the specification block.
    • In the example, the first place-holder in the message-text line:
    <ME01>Page out rate reveals %s %s memory load
    is determined by the value thresholds in the corresponding text-qualifier:
    VM-PAGE-OUT-RATE | 20 an | 15 a | 10 a |
  • The second conversion specifiers in the <message-text> corresponds to the second text qualifier line in the block, and so on.
    • In the example, the first place-holder in the message-text line:
    <ME01>Page out rate reveals %s %s memory load
    is determined by the value thresholds in the corresponding text-qualifier:
    VM-PAGE-OUT-RATE | 20 EXCESSIVE | 15 HEAVY | 10 moderate |
    The item-name <value> determines which <string> text is inserted into the printed advice message.

    Configuration Rules

  • Comment lines must be preceded by a number sign character (#).
  • The first line of the SOS advice file is followed by any number of user-notification commands, terminated by one or more blank lines.
    • User-notification commands can be used to redirect copies of advice messages to another output device. No validation is done to confirm the syntax of these lines. The actual message text should not be included in the command. Instead, the advice message will be appended to the end of it.
    Example
    echo>/dev/console7
  • The rest of the file contains any number of message advice specification blocks separated by one or more blank lines. Each advice specification block must contain a message-id code followed by the actual advice message-text on the first line. Subsequent lines contain threshold criteria.
  • The message-id code is made up of the following components:
  • A type code, which denotes the specific system activity monitored.
  • B for buffer cache activity
  • C for CPU activity
  • D for disk activity
  • G for global activity
  • M for memory activity
  • L for network activity
  • P for process activity
  • A user-defined priority code assigned to the <variable>
  • I indicates the advice message is informational.
  • E indicates the performance level is exceptional or excessive.
  • A unique two-digit identification number (00-99)
  • The <message-id> code is followed by the message text (<message-text>).
    • Example
    <CE01> The CPU Queue length indicates %s %s CPU bottleneck
    The message identification code precedes the message text in the specification file, but follows the message text in the actual advice message display.
  • Conversion specifications in the <message-text> specification must be introduced by the percent sign character (%). After the % character, a conversion character (either s or %) will indicate the type of conversion to by applied.
  • %s (percent sign followed immediately by a lower-case s) indicates the argument is a string and characters from the string will be printed until the end of the string.
  • %% (percent sign followed immediately by a percent sign) will print a % character; no argument is converted.
    • For each variable text or value to be included in the message text, a single-line text qualifier must follow the basic advice specification.
  • If the advice message should always be displayed, the second line of the advice specification block can be replaced with the word ALWAYS to specify the message should always be generated. The <item-name> from the .itemlist file would then be the only entry on the third line of the block.
    • Example
    <PI01>This interval’s ’hog’ process is %s with %s%% of the CPU
    ALWAYS
    %CPUPCT-PID
    %CPUPCT
  • The item-name specification used to determine the text string is usually, but not necessarily, the same as the advice threshold item. An item-name can be selected from block types 0, 6, 7, 8, 10, 12, 14, or 15 in the itemlist file. Or, it can be one of six special item-names preceded by a percent sign (%item-name).
    • The following three items can only be used as variable text item-names. They will be replaced with a string of the form #nnn (nnn=PIN) to identify the appropriate process:
  • %CPU-HOG, which identifies the CPU hog process
  • %DISC-HOG, which identifies the disk hog process
  • %TERM-HOG, which identifies the terminal read hog process
    • The next three special items can be used anywhere as a regular item-name can be used:
  • %HOG-CPU, the CPU percentage used by %CPU_HOG
  • %HOG-DISK, the disk I/O’s performed by %DISC_HOG
  • %HOG-TERM, the terminal reads performed by %TERM_HOG
  • An item-name preceded by an exclamation character (!item-name), specifies that all occurrences of this advice message will be sent through user-notification commands.

    • SNMP Traps

    SOS Performance Advisor provides the ability to send SNMP (Simple Network Management Protocol) traps to an SNMP event browser, such as OpenView Network Node Manager Alarm Browser. The executable program used to accomplish this, snmptrap, comes with the event browser—it is not shipped with the SOS Performance Advisor product.

    Installing the lpstrap File

    Before enabling SNMP traps, you must first install the lpstrap file on your host system.
  • Save the following file as /opt/lps/bin/lpstrap on your host system:

    • #!/bin/ksh -f

    MGR_HOST=<host-systemname>

    SNMPTRAP_PATH=/opt/OV/bin

    $SNMPTRAP_PATH/snmptrap "" .1.3.6.1.4.1.11.2.17.1

    $MGR_HOST 6 58916872\""\

    .1.3.6.1.4.1.11.2.17.2.1.0 Integer 14 \

    .1.3.6.1.4.1.11.2.17.2.5.0 octetstringascii "Major" \

    .1.3.6.1.4.1.11.2.17.2.4.0 octetstringascii \

    "SOS Performance Advisor: $@"
  • Change the file permissions as executable:
    • chmod 755 lpstrap

    Enabling the SNMP Traps

    To enable SNMP traps, perform the following steps.
  • Modify MGR_HOST in /opt/lps/bin/lpstrap to reflect the host that will receive the traps (the system running the browser).
  • Modify SNMPTRAP_PATH in /opt/lps/bin/lpstrap to reflect the path for snmptrap on the host executing SOS Performance Advisor. By default, lpstrap uses /opt/OV/bin/.
  • Modify the /etc/opt/lps/cfg/advice file to enable lpstrap by removing the number sign character (#) in the line: #lpstrap.
  • Modify the /etc/opt/lps/cfg/advice file to specify which messages you wish to be sent as SNMP traps by preceding the threshold specification with a greater than sign (>).
    • For example, the advice message specification block:
    <CE01>CPU Queue length indicates %s %s CPU bottleneck
    CPU-QUEUE-LEN (5-9999)
    CPU-QUEUE-LEN | 10 an | 5 a | 2 a
    CPU-QUEUE-LEN | 10 EXCESSIVE | 5 HEAVY | 2 moderate
    will become:
    <CE01>CPU Queue length indicates %s %s CPU bottleneck
    >CPU-QUEUE-LEN (5-9999)
    CPU-QUEUE-LEN | 10 an | 5 a | 2 a
    CPU-QUEUE-LEN | 10 EXCESSIVE | 5 HEAVY | 2 moderate

    NOTE Although you can enable traps for all advice messages, this feature was designed to notify personnel of exceptional performance levels. For instance, enabling an SNMP trap for an advice message that is ALWAYS generated could be excessive and is not recommended

  • Start the SOS Performance Advisor executable program (SOS or SOSLOGD) to which you want to send the traps, and enable advice messages within that program.
  • For instructions to enable advice messages in SOS, refer to "Display advice messages".
  • For instructions to enable advice messages in SOSLOGD, see "Setting Advanced Configuration Parameters".

    • NOTE If advice messages are enabled in more than one executable program or more than one occurrence of the same program, each program will create SNMP traps. To avoid duplication, enable the advice messages in SOSLOGD only.

    SOS holidays File

    The /etc/opt/lps/cfg/holidays file contains a list of dates to be ignored by SOSLOGX. By default, the file contains exclusion dates for the following holidays in the years1996 through 2010:
  • New Years Day (January 1)
  • Presidents Day (3rd Monday in February)
  • Memorial Day (last Monday in May)
  • Independence Day (July 4)
  • Labor Day (1st Monday in September)
  • Veterans’ Day (November 11)
  • Thanksgiving Day (4th Thursday in November)
  • Christmas Day (December 25)
    • The portion of the /etc/opt/lps/cfg/holidays file that excludes holidays for the year 2000 is provided as an example:
    ! 2000 Holidays
    !
    01/01/00 New Year’s
    02/21/00 President’s Day
    05/29/00 Memorial Day
    07/04/00 Independence Day
    09/04/00 Labor Day
    11/11/00 Veteran’s Day
    11/23/00 Thanksgiving
    12/25/00 Christmas
    Figure 8.2 SOS holidays configuration file (example)
    The purpose of the holidays file is to eliminate atypical computer performance data from the statistical analysis done by SOSLOGX. To add, delete, or modify the contents of this file, use the configuration rules listed below.

    Configuration Rules

    When you know in advance that computer resources used on particular date will not be typical and don’t want that day’s performance to skew performance statistics, you can exclude that date from SOSLOGX’s computations by doing the following:
  • Add the date to the /etc/opt/lps/cfg/holidays file.
  • Use the format MM/DD/YY.
  • Precede any comment lines with an exclamation character (!).
  • Enable Exclusions in SOSLOGX.
  • Enable Holiday Exclusions in SOSLOGX.

    • SOS ppoints File

    The /etc/opt/lps/cfg/ppoints file contains the configuration information for the Pulse Points screen. For information about pulse points, see "SOS Pulse Points Summary"
     
     
    ###############################################################################
    CPU Pulse Points Indicator Lines
    —
    $PP_CPU CPU-BUSY% "CPU Busy %" 60,85 ""
     
    $PP_CPU CPU-HIGH-PRI-BUSY% "Hi-Pri CPU %" 60,85 ""
     
    $PP_CPU CPU-QUEUE-BUSY% "Queue Busy %" 75,90 ""
     
     
    $PP_CPU CPU-QUEUE-LEN "Run-Q Average" 5,10 ""
     
    Memory Pulse Points Indicator Lines
    —
    $PP_MEMORY VM-PAGE-OUT-RATE "Page Out Rate" 15,20 "/sec"
     
    $PP_MEMORY VM-DEACT-BPS "Deactivate Byte Rate" 1, 200 "bytes/sec"
     
    $PP_MEMORY CPU-VFLT% "CPU Memory Mgt %" 3, 7 ""
     
     
    $PP_MEMORY BC-RHIT% "Read Hit %" 90,80 ""
     
    Disk Pulse Points Indicator Lines
    —
    $PP_DISC DISC-AVG-WAIT-TIME "Average Wait Time" 30,40 "System Wide"
     
    $PP_DISC DISC-QUEUE-LEN "Average Q-Length" 1, 3 "System Wide"
     
    $PP_DISC DISC-UTIL% "Disk Utilization %" 40,60 "System Wide"
     
     
    $PP_DISC DISC-IO-RATE "Disk I/O Rate (/sec)" 40,60 "System Wide"
     
    Network Pulse Points Indicator Line
    —
    $PP_NET NETIF-COLLISION% "Collision %" 15,30 "System Wide"
    .
    Figure 8.3 SOS ppoints configuration file (example)

    SOS ppoints File Configuration

    An example of the Pulse Points screen is shown in Figure 8.4.


    Figure 8.4 SOS Pulse Points screen (example)
    By default, the pulse point thresholds and messages are configured for you. You can edit the /etc/ opt/lps/cfg/ppoints file in order to:
  • Add, delete, or reorder the pulse point indicators (variables) that appear in each section
  • Modify the Green (normal), Yellow (problematic), and Red (unacceptable) threshold values
  • Modify the comments associated with each pulse point indicator

    • Pulse Point Indicator Lines

    Example
    $PP_MEMORY VM-DEACT-BPS "Deactivate Byte Rate" 1, 200 "bytes/sec"
    Syntax
    <section><value-spec><label><yellow-threshold, red-threshold><comment>

    Configuration Rules

    Use the following configuration rules when editing the ppoints file.
  • Any pulse points variable that you want to display in the Pulse Points screen must be defined in the /etc/opt/lps/cfg/ppoints file.
  • The first four specification fields in the pulse points indicator line must be completed. The <comments> field may be omitted.
  • Commas, spaces, or tabs must separate the specification fields in the pulse points indicator line to allow for "white space" in the display.
  • Each indicator line must begin with the name of the section in which the variable will appear in the Pulse Points screen. The section name in the <section> field must be preceded by "$PP_". The valid section names are: