Opens the Performance Monitor to display, on a character-cell
terminal, the usage statistics for a database. See the Oracle
Rdb7 Guide to Database Performance and Tuning for tutorial
information on how to interpret the Performance Monitor displays.
1 – Description
The Performance Monitor dynamically samples activity statistics
on a database. You can display the statistics at your terminal
and can also write them to a formatted binary file.
The statistics show activity only from the node on which you
execute the command.
The Performance Monitor operates in one of three modes: online,
record, and replay. In online mode, you can display or record
current activity on a database. In record mode, you can record
statistics in a binary file. In replay mode, you can examine a
previously recorded binary statistics file.
If you use the Input qualifier, the Performance Monitor executes
in replay mode. In replay mode, this command generates an
interactive display from a previously recorded binary statistics
file.
If you do not use the Input qualifier, you must specify a
database file name. The Performance Monitor then executes in
online mode. In online mode, the command generates an interactive
display when you use the Interactive qualifier and can also
record statistics in a binary file.
The interactive display is made up of numerous output pages.
You control the interactive display by means of menus, arrow
keys, and the Return key to select options. You select an item
by pressing the arrow keys until the desired item is highlighted,
then press the Return key.
Display the Select Display options (by typing D) from the
Performance Monitor screen to view the available output pages.
Items in the Display menu followed by this set of characters:
[->, indicate that a submenu is displayed when you select this
item.
Once you have selected a display, there are a number of methods
you can use to navigate through the screens:
o To move to the next screen of information, do one of the
following:
- Press the right arrow (- > ) keyboard key.
- Press the Next Screen keyboard key.
o To move to the previous screen of information, do one of the
following:
- Press the left arrow (< - ) keyboard key.
- Press the Prev Screen keyboard key.
o To move forward n number of screens, press the plus (+)
keyboard key and enter the value n.
o To move backward n number of screens, press the minus (-)
keyboard key and enter the value n.
o To move directly from the first screen to the last screen, do
one of the following:
- Press the up arrow (^ ) keyboard key.
- Press the plus (+) keyboard key and enter the value 0.
o To move directly from the last screen to the first screen, do
one of the following:
- Press the down arrow (v ) keyboard key.
- Press the hyphen (-) keyboard key and enter the value 0.
o To quickly locate a screen in the current submenu group that
contains activity, press the space bar on your keyboard.
This feature works even when you are replaying a binary input
file. If there is no screen in the current subgroup that has
activity, the next screen is displayed (as though you had
used the Next Screen key). The Performance Monitor ignores
computational screens, such as Stall Messages, Monitor Log,
and so on, when searching for activity.
In interactive mode, enter an exclamation point to open the
Select Tool menu. This menu allows you to switch the database
for which you are displaying statistics, edit a file, invoke a
system command, and so on. (The ability to open a new database
is not available if you specify the Input or Output qualifier.)
In addition, it provides you the ability to locate a specific
statistics screen either by name (or portion thereof) or by a
summary-selection menu. Select the Goto screen or Goto screen
"by-name" options from the Select Tool menu to use these options.
In interactive mode, you can pause output scrolling on your
screen by pressing the P key. Resume output scrolling by pressing
the P key again.
An extensive online help facility for the character-cell
interface is available by doing the following from the
Performance Monitor screen:
1. Type H or PF2.
2. Select the type of help you want (keyboard, screen, or field).
3. Press the Return key.
If you select field level help, you must also do the following:
1. Highlight the field for which you want help information.
2. Press the Return key.
All screens regardless of format or display contents have a
standard format as follows:
o First line
Contains the node name, the utility name and version number,
and the current system date and time. The current system date
and time are updated at the specified set-rate interval.
o Second line
Contains the screen refresh rate, in seconds; the current
screen name; and the elapsed time since the last set-rate
command, which indicates how long the screen information has
been collected.
o Third line
Contains the current page number within the screen (screen X
of Y), the name of the current database, and the statistics
utility operation mode (online, record, or replay). Online
mode is the normal database activity displayed in real
time. Record mode indicates that the database activity being
displayed is being recorded to an external file specified by
the Output qualifier. Replay mode indicates that the database
activity is being displayed from the external file specified
by the Input qualifier.
You can display most statistics in either a histogram or a
columnar chart, although several display pages have special
formats. By default, the initial interactive display appears
in histogram mode; by using the Nohistogram qualifier, you can
direct Oracle RMU to display statistics in tabular numeric mode.
In addition, you can produce time-plot graphics for individual
statistical fields.
Use the Output qualifier to direct statistical output to a file.
The output is a formatted binary file and does not produce a
legible printed listing. To read the output, you must use the RMU
Show Statistics command with the Input qualifier.
The Nointeractive qualifier suppresses the interactive display.
Use this qualifier when you want to generate binary statistics
output but do not want an online display.
Database statistics are maintained in a global section on each
system on which Oracle Rdb is running. Statistics are reset to
zero when you close a database. Running the Performance Monitor
keeps the database open even when there are no users accessing
the database.
The Stall Messages display permits you to display multiple
screens of information. Access the Stall Messages display by
selecting Per-Process Information from the Select Display Menu;
then select the Stall Messages display from the secondary menu.
If you are displaying the last screen of Stall Messages
information and the number of stalled processes is reduced such
that the last screen is empty, you are automatically moved to the
newest last screen of information when you press the Next Screen
keyboard key (or the right arrow keyboard key).
You can also use the Alarm, Notify, and Screen qualifiers to
simplify monitoring stalled processes. See the description of
each of these qualifiers for more information.
2 – Format
(B)0[m RMU/Show Statistics [root-file-spec]
[4mCommand[m [4mQualifiers[m x [4mDefaults[m
x
/Access_Log x None
/Alarm=interval x /Alarm=0
/[No]Broadcast x See description
/[No]Cluster=[(node-list)] x /Nocluster
/Configure=file-spec x None
/[No]Cycle=seconds x /Nocycle
/Dbkey_Log=file-spec x See description
/Deadlock_Log=file-spec x None
/[No]Histogram x /Histogram
/Hot_Standby_Log x None
/Input = file-name x See description
/[No]Interactive x See description
/Lock_Timeout_Log=file-spec x None
/[No]Log x See description
/[No]Logical_Area x /Logical_Area
/[No]Notify[=([No]All | operator-classes)] x /Nonotify
/[No]Opcom_Log=filename x /Noopcom_Log
(B)0[m /Options=keywords x /Options=Base
/Output=file-spec x See description
/[No]Prompt_Timeout=seconds x /Prompt_Timeout=60
/Reopen_Interval= minutes x None
/Reset x Statistics are no
/Screen = screen-name x See description
/Stall_Log = file-spec x Stall messages no
/Time = integer x /Time = 3
/Until = date-time x See description
3 – Parameters
3.1 – root-file-spec
The root file specification of the database on which you
want statistics. If you use the Input qualifier to supply a
prerecorded binary statistics file, you cannot specify a database
file name. If you do not use the Input qualifier, you must
specify a database file name.
4 – Command Qualifiers
4.1 – Access Log
Identifies the name of the log file where logical area accesses
are to be recorded.
4.2 – Alarm
Alarm=interval
Establishes an alarm interval (in seconds) for the Stall Messages
screen from the command line. This is useful when you plan to
submit the RMU Show Statistics command as a batch job.
Use this qualifier in conjunction with the Notify qualifier to
notify an operator or set of operators of stalled processes.
The default value is 0 seconds, which is equivalent to disabling
notification.
4.3 – Broadcast
Broadcast
Nobroadcast
Specifies whether or not to broadcast messages. The Broadcast
qualifier is the default, if broadcasting of certain messages
has been enabled with DCL SET BROADCAST. If broadcasting has
been disabled with the DCL SET BROADCAST=none command, broadcast
messages are not displayed, even if you specify the RMU Show
Statistics command with the Broadcast qualifier.
Specify the Nobroadcast qualifier if broadcasting has been
enabled with the DCL SET BROADCAST command but you do not
want broadcast messages displayed while you are running the
Performance Monitor.
4.4 – Cluster
Cluster=(node-list)
Nocluster
Specifies the list of remote nodes from which statistics
collection and presentation are to be performed. The collected
statistics are merged with the information for the current node
and displayed using the usual statistics screens.
The following list summarizes usage of the Cluster qualifier:
o If the Cluster qualifier is specified by itself, remote
statistics collection is performed on all cluster nodes on
which the database is currently open.
o If the Cluster=(node-list) qualifier is specified, remote
statistics collection is performed on the specified nodes
only, even if the database is not yet open on those nodes.
o If the Cluster qualifier is not specified, or the Nocluster
qualifier (the default) is specified, cluster statistics
collection is not performed. However, you can still enable
clusterwide statistics collection online using the Tools menu.
You can specify up to 95 different cluster nodes with the Cluster
qualifier. There is a maximum number of 95 cluster nodes because
Oracle Rdb supports only 96 nodes per database. The current node
is always included in the list of nodes from which statistics
collection is to be performed.
It is not necessary to have the RMU Show Statistics command
running on the specified remote nodes or to have the database
open on the remote nodes. These events are automatically handled
by the feature.
The following example shows the use of the Cluster qualifier to
initiate statistics collection and presentation from two remote
nodes:
$ RMU /SHOW STATISTICS /CLUSTER=(BONZAI, ALPHA4) MF_PERSONNEL
Remote nodes can also be added and removed online at run time.
Use the Cluster Statistics option located in the Tools menu.
The Tools menu is displayed by using the exclamation point (!)
on-screen menu option.
See the RMU Show Statistic DBA Handbook (available in MetaLink
if you have a service contract) for information about the Cluster
Statistics Collection and Presentation feature.
4.5 – Configure
Configure=file-spec
Specifies the name of a human-readable configuration file to be
processed by the RMU Show Statistics command. The configuration
file can be created using any editor, or it can be automatically
generated from the RMU Show Statistics command using the current
run-time configuration settings. The default configuration file
type is .cfg.
If you specify the Configure=file-spec qualifier, the
configuration file is processed by the RMU Show Statistics
command prior to opening the database or the binary input file.
If you do not specify this qualifier, all of the variables are
the defaults based on command-line qualifiers and logical names.
The configuration file is processed in two passes. The first
pass occurs before the database is opened and processes most
of the configuration file entries. The second pass occurs after
the database is opened and processes those variables that are
database-dependent, such as the CUSTOMER_LINE_n variable.
See the RMU Show Statistic DBA Handbook (available in MetaLink
if you have a service contract) for more information about
configuration files.
4.6 – Cycle
Cycle=seconds
Nocycle
Directs the Performance Monitor to continually cycle through the
set of screens associated with the currently selected menu item.
Each menu is displayed for the number of seconds specified.
When you specify the Cycle qualifier, you can change screen
modes or change submenus as desired; cycling through the menus
associated with your choice continues at whichever menu level is
currently selected.
The specified value for the Cycle qualifier must be greater
than or equal to the value specified for the Time qualifier.
In addition, if you manually change the refresh rate (using the
Set_rate onscreen menu option) to a value that is greater than
the value you specify with the Cycle qualifier, the cycling is
performed at the interval you specify for the Set_rate.
If you do not specify the Cycle qualifier, or if you do not
specify the number of seconds, no screen cycling is performed.
4.7 – Dbkey Log
Dbkey_Log=file-spec
Logs the records accessed during a given processing period by the
various attached processes. The file-spec is the name of the file
to which all accessed dbkeys are logged.
The header region of the dbkey log contains four lines. The first
line indicates that the RMU Show Statistic utility created the
log file. The second line identifies the database. The third
line identifies the date and time the dbkey log was created. The
fourth line is the column heading line.
The main body of the dbkey log contains six columns. The first
column contains the dbkey process ID and stream ID. The second
through sixth columns contain the most recently accessed dbkey
for a data page, snapshot page, SPAM page, AIP page, and ABM
page, respectively.
Only one message per newly accessed dbkey is recorded. However,
all dbkey values are displayed, even if some of the dbkeys did
not change.
The dbkey information is written at the current screen refresh
rate, determined by the Time qualifier or the Set_rate onscreen
menu option. Using a larger refresh rate minimizes the size of
the file but results in a large number of missed dbkey messages.
Using a smaller refresh rate produces a large log file, but
contains a much finer granularity of dbkey messages.
Note that you do not need to display the Dbkey Information screen
in order to record the dbkey messages to the dbkey log. The
dbkey log is maintained regardless of which screen, if any, is
displayed.
You can use the Dbkey_Log qualifier to construct a dbkey logging
server, as follows:
$ RMU/SHOW STATISTICS/NOHISTOGRAM/TIME=1 -
_$ /NOINTERACTIVE/DBKEY_LOG=DBKEY.LOG MF_PERSONNEL -
_$ /NOBROADCAST/UNTIL="15:15:00"
4.8 – Deadlock Log
Deadlock_Log=file-spec
Records the last deadlock for the processes. There is no method
to record each lock deadlock as it occurs.
The file-spec in the qualifier is the name of the file to which
you want all lock deadlock messages to be logged. The lock
deadlock messages are written in human-readable format similar
to the Lock Timeout History and Lock Deadlock History screens.
The header region of the lock deadlock log contains three lines:
o Line 1 indicates that the RMU Show Statistics utility created
the log file.
o Line 2 identifies the database.
o Line 3 identifies the date and time the log was created.
The main body of the stall log contains three columns:
o The first column contains the process ID and stream ID that
experienced the lock deadlock.
o The second column contains the time the deadlock occurred;
however, the date is not displayed.
o The third column contains the deadlock message describing the
affected resource. This message is similar to the originating
stall message.
For example:
2EA00B52:34 14:25:46.14 - waiting for page 5:751 (PR)
If any lock deadlocks are missed for a particular process
(usually because the recording interval is too large), the
number of missed lock deadlocks is displayed in brackets after
the message. For example:
2EA00B52:34 14:25:46.14 - waiting for page 5:751 (PR) [1 missed]
Only one message is logged for each deadlock.
The lock deadlock messages are written at the specified screen
refresh rate, determined by specifying the Time qualifier, or
online using the Set_rate on-screen menu option. Using a larger
refresh rate minimizes the size of the file, but results in
a large number of missed deadlock messages. Using a smaller
refresh rate produces a large log file, but contains a much finer
granularity of deadlock messages.
Using the Time=1 or Time=50 qualifier produces a reasonable log
while minimizing the impact on the system.
The affected LockID is not displayed, because this is meaningless
information after the lock deadlock has completed.
Use the Tools menu (displayed when you press the exclamation
point (!) key from any screen) to enable or disable the lock
timeout and lock deadlock logging facility while the RMU Show
Statistics utility is running. However, note that the lock
timeout log and lock deadlock log are not available during binary
file replay.
4.9 – Histogram
Nohistogram
Directs Oracle RMU to display the initial statistics screen in
the numbers display mode or the graph display mode. The Histogram
qualifier specifies the graph display mode. The Nohistogram
qualifier specifies the numbers display mode.
The Histogram qualifier is the default.
4.10 – Hot Standby Log
Specifies the name of the Hot Standby log file. The "Start hot
standby logging" option of the Tools menu (enter !) can be used
to specify the name of the Hot Standby log file at runtime.
4.11 – Input
Input=file-name
Specifies the prerecorded binary file from which you can read the
statistics. This file must have been created by an earlier RMU
Show Statistics session that specified the Output qualifier.
You cannot specify a database file name with the Input qualifier.
Also, you must not use the Until, Output, or Nointeractive
qualifiers with the Input qualifier. However, you can use the
Time qualifier to change the rate of the display. This will not
change the computed times as recorded in the original session.
For example, you can record a session at Time=60. This session
will gather statistics once per minute.
You can replay statistics gathered in a file by using the Input
and Time qualifiers. To replay a file:
o Use the Output qualifier to create a file of database
statistics.
o Use the Input and Time qualifiers to view the statistics
again at a rate that you determine. For example, the command
RMU/SHOW STATISTICS PERS.LOG/TIME=1, will replay the PERS.LOG
file and change the display once per second, thus replaying 10
hours of statistics in 10 minutes.
If you do not specify the Input qualifier, you must specify the
root-file-spec parameter.
4.12 – Interactive
Nointeractive
Displays the statistics dynamically to your terminal. The
Interactive qualifier is the default when you execute the
RMU Show Statistics command from a terminal. You can use the
Nointeractive qualifier with the Output qualifier to generate a
binary statistics file without generating a terminal display. The
Nointeractive qualifier is the default when you execute the RMU
Show Statistics command from a batch job.
In an interactive session, you can use either the menu interface
or the predefined control characters to select display options
(see the Performance Monitor online help for further information
about the predefined control characters).
Select menu options by using the up (^ ) and down (v ) arrow keys
followed by pressing the Return or Enter key. Cancel the menu by
pressing Ctrl/Z.
4.13 – Lock Timeout Log
Lock_Timeout_Log=file-spec
Records the last lock timeout message for the processes. There
is no method to record each lock timeout as it occurs. The lock
timeout messages are written in human-readable format.
The header region of the lock timeout log contains three lines:
o Line 1 indicates that the RMU Show Statistics utility created
the log file.
o Line 2 identifies the database.
o Line 3 identifies the date and time the log was created.
The main body of the stall log contains three columns:
o The first column contains the process ID and stream ID that
experienced the lock timeout.
o The second column contains the time the timeout occurred;
however, the date is not displayed.
o The third column contains the timeout message describing the
affected resource. This message is similar to the originating
stall message.
For example:
2EA00B52:34 14:25:46.14 - waiting for page 5:751 (PR)
If any lock timeouts are missed for a particular process (usually
because the recording interval is too large), the number of
missed lock timeouts is displayed in brackets after the message.
For example:
2EA00B52:34 14:25:46.14 - waiting for page 5:751 (PR) [1 missed]
Only one message is logged for each lock timeout.
The lock timeout messages are written at the specified screen
refresh rate, determined by specifying the Time qualifier, or
online using the Set_rate on-screen menu option. Using a larger
refresh rate minimizes the size of the file, but results in a
large number of missed lock timeout messages. Using a smaller
refresh rate produces a large log file, but contains a much finer
granularity of lock timeout messages.
Using the Time=1 or Time=50 qualifier appears to produce a
reasonable log while minimizing the impact on the system.
The affected LockID is not displayed because this is meaningless
information after the lock timeout has completed.
Note that you do not need to be displaying the Lock Timeout
History or Lock Deadlock History screens to record the stall
messages to the stall log. These logs are maintained regardless
of which screen, if any, is displayed.
Use the Tools menu (displayed when you press the exclamation
point (!) key from any screen) to enable or disable the lock
timeout and lock deadlock logging facility while the RMU Show
Statistics utility is running. However, note that the lock
timeout log and lock deadlock log are not available during binary
file replay.
4.14 – Log
Nolog
Logs the creation of a binary statistics file to your output
file. This binary statistics file is created only if you have
used the Output qualifier. If you use the Nolog qualifier, no
operations will be logged to your output file.
The default is the current setting of the DCL verify switch. See
HELP SET VERIFY in DCL HELP for more information on changing the
DCL verify switch.
If you use the Interactive qualifier, the Log qualifier is
ignored.
4.15 – Logical Area
Logical_Area
Nological_Area
Specifies that you want the RMU Show Statistics command to
acquire the needed amounts of virtual memory to display logical
area statistics information. The Logical_Area qualifier is the
default.
By default, the RMU Show Statistics command consumes
approximately 13,000 bytes of virtual memory per logical area.
(The number of logical areas is determined by the largest logical
area identifier - not by the actual number of areas.) This can
result in the RMU Show Statistics command consuming large amounts
of virtual memory, even if you do not want to review logical area
statistics information.
Use the NoLogical_Area qualifier to indicate that you do not want
to display logical area statistics information. When you specify
the NoLogical_Area qualifier, the virtual memory for logical area
statistics information presentation is not acquired.
When you specify the NoLogical_Area qualifier, do not also
specify the Nolog qualifier, as this causes logical area
statistics information to still be collected.
The "Logical Area" statistics are not written to the binary
output file. Conversely, the "Logical Area" statistics screens
are not available during binary input file replay.
There is no corresponding configuration variable. This qualifier
cannot be modified at run time. See the RMU Show Statistic DBA
Handbook (available in MetaLink if you have a service contract)
for more information about interpreting logical area screens.
4.16 – Notify
Notify
Notify=All
Notify=Noall
Notify=operator-classes
Nonotify
Notifies the specified system operator or operators when a
stall process exceeds the specified alarm interval by issuing
a broadcast message and ringing a bell at the terminal receiving
the message.
The valid operator classes are: CENTRAL, CLUSTER, DISKS, OPCOM,
SECURITY, and OPER1 through OPER12.
The various forms of the Notify qualifier have the following
effects:
o If you specify the Notify qualifier without the operator-
classes parameter, the CENTRAL and CLUSTER operators are
notified by default.
o If you specify the Nonotify or Notify=Noall qualifiers,
operator notification is disabled.
o If you specify the Notify=All qualifier, all operator classes
are enabled.
o If you specify the Notify=operator-classes qualifier, the
specified classes are enabled. (If you specify more than one
operator class, enclose the list in parentheses and separate
each class name with a comma.)
For example, issuing the RMU Show Statistics command with the
Notify=(OPER1, OPER2) qualifier sends a notification message
to system operator classes OPER1 and OPER2 if the Alarm
threshold is exceeded while monitoring the Stall Messages
screen.
o When the Notify=OPCOM qualifier is specified with the RMU
Show Statistics command along with the Alarm and Cluster
qualifiers, Oracle RMU generates an OPCOM message and delivers
it to the OPCOM class associated with the Notify qualifier.
This message alerts the operator to the fact that the process
has stalled for more than n seconds, where n is the value
assigned to the Alarm qualifier. The process that has stalled
may be on any node that is included in the node name list
assigned to the Cluster qualifier.
The specified system operator(s) are notified only when the alarm
threshold is first exceeded. For instance, if three processes
exceed the alarm threshold, the specified operator(s) are
notified only once. If another process subsequently exceeds the
alarm threshold while the other processes are still displayed,
the specified system operator(s) are not notified.
However, if the longest-duration stall is resolved and a new
process then becomes the newest stall to exceed the alarm
threshold, then the specified system operator(s) will be notified
of the new process.
To receive operator notification messages, the following three
OpenVMS DCL commands must be issued:
1. $ SET TERM /BROADCAST
2. $ SET BROADCAST=OPCOM
3. $ REPLY /ENABLE=(operator-classes)
The operator-classes specified in the REPLY /ENABLE command must
match those specified in the Notify qualifier to the RMU Show
Statistics command.
The operator notification message will appear similar to the
following sample message:
%%%%%%%%%%% OPCOM 19-DEC-1994 08:56:39.27 %%%%%%%%%%%
(from node MYNODE at 19-DEC-1994 08:56:39.30)
Message from user SMITH on MYNODE
Rdb Database USER2:[SMITH.WORK.AIJ]MF_PERSONNEL.RDB;1 Event Notification
Process 2082005F:1 exceeded 5 second stall: waiting for record 51:60:2 (EX)
The system operator notification message contains four lines.
Line 1 contains the OPCOM broadcast header message. Line 2
identifies the process running the RMU Show Statistics command
that sent the message. Line 3 identifies the database being
monitored. Line 4 identifies the process that triggered the
alarm, including the alarm interval and the stall message.
To establish an alarm interval for the Stall Messages screen, use
the Alarm=Interval qualifier.
If you specify the Nointeractive qualifier, bell notification is
disabled, but the broadcast message remains enabled.
4.17 – Opcom Log
Opcom_Log=filename
Noopcom
Specifies the name of the file where OPCOM messages broadcast by
attached database processes will be sent.
When recording OPCOM messages, it is possible to occasionally
miss a few messages for a specific process. When this occurs, the
message "n missed" will be displayed in the log file.
You can record specific operator classes of OPCOM messages if
you specify the Option=Verbose qualifier. The Option=Verbose
qualifier records only those messages that can be received by the
process executing the RMU Show Statistics utility. For example,
if the process is enabled to receive operator class Central, then
if you specify Opcom_Log=opcom.log the Option=Verbose qualifier
records all Central operator messages. Conversely, specifying
only the Opcom_Log=opcom.log qualifier records all database-
specific OPCOM messages generated from this node. Because the
output is captured directly from OpenVMS, the operator-specific
log file output format is different from the database-specific
contents. The following example shows the operator-specific log
file contents for the Cluster and Central operator classes:
Oracle Rdb X7.1-00 Performance Monitor OPCOM Log
Database KODA_TEST:[R_ANDERSON.TCS_MASTER]TCS.RDB;2
OPCOM Log created 11-JUN-1999 10:52:07.53
11-JUN-1999 10:52:23.85) Message from user RDBVMS on ALPHA4 Oracle Rdb X7.1-00
Event Notification for Database _$111$DUA368:[BBENTON.TEST]MF_PERSONNEL.RDB;1
AIJ Log Server terminated
11-JUN-1999 10:52:25.49) Message from user RDBVMS on ALPHA4 Oracle Rdb X7.1-00
Event Notification for Database _$111$DUA368:[BBENTON.TEST]MF_PERSONNEL.RDB;1
AIJ Log Roll-Forward Server started
11-JUN-1999 10:52:26.06) Message from user RDBVMS on ALPHA4 Oracle Rdb X7.1-00
Event Notification for Database _$111$DUA368:[BBENTON.TEST]MF_PERSONNEL.RDB;1
AIJ Log Roll-Forward Server failed
.
.
.
11-JUN-1999 10:54:21.09) Message from user RDBVMS on ALPHA4 Oracle Rdb X7.1-00
Event Notification for Database _$111$DUA368:[BBENTON.TEST.JUNK]T_
PERSONNEL.RDB;1 AIJ Log Server started
11-JUN-1999 10:54:21.13) Message from user RDBVMS on ALPHA4 Oracle Rdb X7.1-00
Event Notification for Database _$111$DUA368:[BBENTON.TEST.JUNK]T_
PERSONNEL.RDB;1 Opening "$111$DUA368:[BBENTON.TEST.JUNK]TEST1.AIJ;2"
4.18 – Options
The following keywords may be used with the Options qualifier:
o [No]All
Indicates whether or not all collectible statistics (all
statistics for all areas) are to be collected. The All option
indicates that all statistics information is to be collected;
the Noall keyword indicates that only the base statistics
information is to be collected. You must also specify the
Output qualifier. Note: Logical Area information is not
written to the binary output file.
o [No]Area
Indicates whether or not the by-area statistics information
is to be collected in addition to the base statistics
information. When you specify the Area or Noarea option, the
Base statistics are implicitly selected. You must also specify
the Output qualifier.
When the Area option is specified, statistics for all existing
storage areas are written to the binary output file; you
cannot selectively choose specific storage areas for which
statistic information is to be collected.
The size of the by-area statistics output largely depends on
the total number of storage areas in the database, including
reserved storage areas. If the database contains a large
number of storage areas, it may not be advisable to use the
Options=Area qualifier.
Before you replay a binary output file that contains by-
area statistics, specify the following command to format the
display correctly:
$ SET TERM/NOTAB
You can then replay the statistics as follows:
$ RMU/SHOW STATISTICS/INPUT=main.stats
o Base (default)
Indicates that only the base set of statistics is to be
collected; this is the default Options option. The base set of
statistics is identical to the one collected prior to Oracle
Rdb V6.1. You must also specify the Output qualifier. You
cannot specify Nobase.
o Compress
Compresses the statistics records written to the output
file specified by the Output qualifier. While replaying the
statistics, the RMU Show Statistics command determines if a
record was written using compression or not. If the record was
written using compression it is automatically decompressed.
If compression is used, the resultant binary file can be
read only by the RMU Show Statistics command. The format and
contents of a compressed file are not documented or accessible
to other applications.
o Confirm
Indicates that you wish to confirm before exiting from the
utility. You can also specify the Confirm option in the
configuration file using the CONFIRM_EXIT variable. A value
of TRUE indicates that you want to confirm before exiting the
utility and a value of FALSE (the default) indicates you do
not want to confirm before exiting the utility.
o Log_Stall_Alarm
If Log_Stall_Alarm is present when using the Stall_Log
qualifier to write stall messages to a log file and the
Alarm qualifier to set an alarm interval, only those stalls
exceeding the Alarm specified duration are written to the
stall log output file.
o Log_Stall_Lock
If you use the Stall_Log qualifier to write stall messages to
a log file, use the Nolog_Stall_Lock option to prevent lock
information from being written to the log file. If you use or
omit the Log_Stall_Lock option, lock information is written to
the log file.
o [No]Row_Cache
Indicates that all row cache related screens and features of
the RMU Show Statistics facility are to be displayed. NoRow_
Cache indicates that these features are disabled.
o Screen_Name
Allows you to identify a screen capture by screen name. If you
issue an RMU Show Statistics command with the Options=Screen_
Name qualifier, the screen capture is written to a file that
has the name of the screen with all spaces, brackets, and
slashes replaced by underscores. The file has an extension of
.SCR. For example, if you use the Option=Screen_Name qualifier
and select the Write option on the Screen Transaction
Duration (Read/Write), the screen is written to a file named
TRANSACTION_DURATION_READ_WRITE.SCR.
o Update
Allows you to update fields in the Database Dashboard. See
the Performance Monitor Help or the Oracle Rdb7 Guide to
Database Performance and Tuning for information about using
and updating the Database Dashboard. You must have both the
OpenVMS WORLD and BYPASS privileges to update fields in the
Database Dashboard.
o Verbose
Causes the stall message logging facility to report a stall
message at each interval, even if the stall message has been
previously reported.
NOTE
Use of the Options=Verbose qualifier can result in an
enormous stall messages log file. Ensure that adequate
disk space exists for the log file when you use this
qualifier.
You can enable or disable the stall messages logging Verbose
option at run time by using the Tools menu and pressing the
exclamation point (!) key.
You can also specify the Verbose option in the configuration
file by using the STALL_LOG_VERBOSE variable. Valid keywords
are ENABLED or DISABLED.
Lock information is displayed only once per stall, even in
verbose mode, to minimize the output file size.
4.19 – Output
Output=file-name
Specifies a binary statistics file into which the statistics are
written. Information in the Stall Messages screen is not recorded
in this file, however. The information in the Stall Messages
screen is highly dynamic and thus cannot be replayed using the
Input qualifier.
NOTE
Statistics from the Stall Messages display are not collected
in the binary output file.
For information on the format of the binary output file (which
changed in Oracle Rdb V6.1), see the Oracle Rdb7 Guide to
Database Performance and Tuning.
4.20 – Prompt Timeout
Prompt_Timeout=seconds
Noprompt_Timeout
Allows you to specify the user prompt timeout interval, in
seconds. The default value is 60 seconds.
If you specify the Noprompt_Timeout qualifier or the Prompt_
Timeout=0, the RMU Show Statistics command does not time out any
user prompts. Note that this can cause your database to hang.
NOTE
Oracle Corporation recommends that you do not use the
Noprompt_Timeout qualifier or the Prompt_Timeout= 0
qualifier unless you are certain that prompts will always
be responded to in a timely manner.
If the Prompt_Timeout qualifier is specified with a value greater
than 0 but less than 10 seconds, the value 10 is used. The user
prompt timeout interval can also be specified using the PROMPT_
TIMEOUT configuration variable.
4.21 – Reopen Interval
Reopen_Interval=minutes
After the specified interval, closes the current output file and
opens a new output file without requiring you to exit from the
Performance Monitor. The new output file has the same name as the
previous output file, but the version number is incremented by 1.
This qualifier allows you to view data written to the output file
while the Performance Monitor is running.
If there has been no database activity at the end of the
specified interval, the current output file is not closed and
a new output file is not created.
Be careful not to use the DCL PURGE command inadvertently. Also
note that use of the DCL SET FILE/VERSION_LIMIT command causes
older versions of the output file to be deleted automatically.
Use of the Reopen_Interval qualifier is only valid when you also
specify the Output qualifier.
4.22 – Reset
Specifies that you want the Performance Monitor to reset your
display to zero. The Reset qualifier has the same effect as
selecting the reset option from the interactive screen (except
when you specify the Reset qualifier, values are reset before
being initially displayed).
Note that this qualifier resets the values being displayed to
your output device only, it does not reset the values in the
database global section nor does it affect the data collected in
an output file.
The default behavior of the Performance Monitor is to display
each change in values that has occurred since the database was
opened. To display only the value changes that have occurred
since the Performance Monitor was invoked, specify the Reset
qualifier, or immediately select the on-screen reset option when
statistics are first displayed.
The Reset qualifier does not affect the values that are written
to the binary output file (created when you specify the Output
qualifier). Specify the Reset qualifier when you replay the
output file if you want the replay to display only the change in
values that occurred between the time the Performance Monitor was
invoked (with the Output qualifier) and the monitoring session
ended.
4.23 – Screen
Screen=screen-name
Specifies the first screen to be displayed. This is particularly
useful when you are using the Performance Monitor to
interactively monitor stalled processes. For example, the
following command automatically warns the system operator of
excessive stalls:
$ RMU/SHOW STATISTICS/ALARM=5/NOTIFY=OPER12/SCREEN="Stall Messages" -
_$ MF_PERSONNEL
The following list describes the syntax of the screen-name
argument:
o You can use any unique portion of the desired screen name for
the screen-name argument. For example, the following has the
same results as the preceding example:
$ RMU/SHOW STATISTICS/ALARM=5/NOTIFY=OPER12/SCREEN="Stall" -
_$ MF_PERSONNEL.RDB
o Except with regards to case, whatever unique portion of the
screen you supply must be an exact match to the equivalent
portion of the actual screen name.
For example Screen="Stall" is equivalent to Screen="STALL";
however Screen="Stalled" is not.
o If the specified screen-name does not match any known screen
name, the display starts with the Summary IO Statistics screen
(the default first screen). No error message is produced.
o If the screen name contains spaces, enclose the screen-name in
quotes.
o You can not specify the "by-lock" or "by-area" screens.
If you specify the Nointeractive qualifier, the Screen qualifier
is ignored.
4.24 – Stall Log
Stall_Log=file-spec
Specifies that stall messages are to be written to the specified
file. This can be useful when you notice a great number of stall
messages being generated, but do not have the resources on hand
to immediately investigate and resolve the problem. The file
generated by the Stall_Log qualifier can be reviewed later so
that the problem can be traced and resolved.
The stall messages are written to the file in a format similar to
the Stall Messages screen. Stall messages are written to the file
at the same rate as the screen refresh rate. (The refresh rate
is set with the Time qualifier or from within the Performance
Monitor with the Set_rate on-screen menu option.) Specifying a
large refresh rate minimizes the size of the file, but results
in a large number of missed stall messages. Specifying a small
refresh rate produces a large log file, but contains more of the
stall messages generated.
You do not need to be displaying the Stall Messages screen to
record the stall messages to the log file. The stall log is
maintained regardless of which screen, if any, is displayed.
By default, stall messages are not logged to a file.
4.25 – Time
Time=integer
Specifies the statistics collection interval in seconds. If
you omit this qualifier, a sample collection is made every 3
seconds. The integer has a normal range of 1 to 180 (1 second
to 3 minutes). However, if you specify a negative number for the
Time qualifier, the RMU Show Statistics command interprets the
number as hundredths of a second. For example, Time=-20 specifies
an interval of 20/100 or 1/5 of a second.
If you are running the RMU Show Statistics command interactively,
it updates the screen display at the specified interval.
If you also use the Output qualifier, a binary statistics record
is written to the output file at the specified interval. A
statistics record is not written to this file if no database
activity has occurred since the last record was written.
4.26 – Until
Until="date-time"
Specifies the time the statistics collection ends. When this
point is reached, the RMU Show Statistics command terminates
and control returns to the system command level. When the
RMU Show Statistics command is executed in a batch job, the batch
job terminates at the time specified.
An example of using the Until qualifier follows:
$ DEFINE LIB$DT_INPUT_FORMAT "!MAU !DB, !Y4 !H04:!M0:!S0.!C2"
$ RMU/SHOW STATISTICS /UNTIL="JUNE 16, 1996 17:00:00.00" -
_$ MF_PERSONNEL
This stops execution of the RMU Show Statistics command at 5 P.M.
on June 16, 1996. You can omit the date if you wish to use the
default of today's date.
You can use either an absolute or delta value to specify the data
and time.
If you do not use the Until qualifier, the RMU Show Statistics
command continues until you terminate it manually. In an
interactive session, terminate the command by pressing Ctrl/Z
or by selecting Exit from the menu. When you are running the RMU
Show Statistics command with the Nointeractive qualifier from a
terminal, terminate the command by pressing Ctrl/C or Ctrl/Y and
then selecting Exit. When you are running the RMU Show Statistics
command in a batch job, terminate the command by deleting the
batch job.
5 – Usage Notes
o Refer to the Oracle Rdb7 Guide to Database Performance and
Tuning for complete information about the RMU Show Statistics
command, including information about using formatted binary
output files from the RMU Show Statistics command.
o To use the RMU Show Statistics command for a database, you
must have the RMU$SHOW privilege in the root file ACL for the
database or the OpenVMS SYSPRV, BYPASS, or WORLD privilege.
To use the RMU Show Statistics command to display statistics
about other users, you must have the OpenVMS WORLD privilege.
To use the RMU Show Statistics command to update fields in
the Database Dashboard (specified with the Options=Update
qualifier), you must have both the OpenVMS WORLD and BYPASS
privileges.
o If a database recovery process is underway, you cannot
exit the Performance Monitor using Ctrl/Z or "E" from the
interactive display menu. You must use Ctrl/Y or wait for the
recovery process to complete. Exiting from the Performance
Monitor causes Oracle RMU to request several locks; however,
these locks cannot be granted because the recovery process
stalls all new lock requests until the recovery is complete.
o Since Oracle Rdb V4.1, a number of changes have been made to
the data structures used for the RMU Show Statistics command.
If you are having a problem with an application that accesses
the RMU Show Statistics field structures, recompile your
application with SYS$LIBRARY:RMU$SHOW_STATISTICS.CDO (or
RMU$SHOW_STATISTICSnn.CDO in a multiversion environment, where
nn is the version of Oracle Rdb you are using).
o The Oracle Rdb RMU Show Statistics command displays process
CPU times in excess of 1 day. Because the width of the CPU
time display is limited, the following CPU time display
formats are used:
- For CPU time values less than 1 day: "HH:MM:SS.CC"
- For CPU time values less than 100 days but more than 1 day:
"DD HH:MM"
- For CPU time values more than 100 days: "DDD HH:MM"
o The following caveats apply to the Cluster Statistics
Collection and Presentation feature:
- Up to 95 cluster nodes can be specified. However, use
cluster statistics collection prudently, as the system
overhead in collecting the remote statistics may be
substantial depending on the amount of information being
transmitted on the network.
- Cluster statistics are collected at the specified display
refresh rate. Therefore, set the display refresh rate to
a reasonable rate based on the number of cluster nodes
being collected. The default refresh rate of 3 seconds is
reasonable for most remote collection loads.
- If you specify the Cluster qualifier, the list of cluster
nodes applies to any database accessed during the Show
Statistics session. When you access additional databases
using the Switch Database option, the same cluster nodes
are automatically accessed. However, any nodes that you
added manually using the Cluster Statistics menu are
not automatically added to the new database's remote
collection.
In other words, manually adding and deleting cluster nodes
affects only the current database and does not apply to
any other database that you may have accessed during the
session. For example, when you run the Show Statistics
utility on node ALPHA3 with manually added node BONZAI,
subsequently switching to BONZAI as the current node will
not display cluster statistics from node ALPHA3 unless you
manually add that node. Furthermore, switching back to node
ALPHA3 as the current node loses the previous collection of
node BONZAI because it was manually added.
- Both DECnet and TCP/IP network protocols are supported.
By default, the DECnet protocol is used. To explicitly
specify which network protocol to use, define the RDM$BIND_
STT_NETWORK_TRANSPORT to DECNET or TCPIP respectively. The
RDM$BIND_STT_NETWORK_TRANSPORT logical name must be defined
to the same definition on both the local and cluster nodes.
The RDM$BIND_STT_NETWORK_TRANSPORT logical name can be
specified in LNM$FILE_DEV on the local node but must be
specified in the LNM$SYSTEM_TABLE on all remote nodes.
NOTE
There is no command qualifier to specify the network
protocol.
- The Output qualifier continues to work as usual, but when
in cluster mode writes the cluster statistics information
to the binary output file.
- The Cluster qualifier cannot be specified with the Input
qualifier. Furthermore, the online selection of cluster
nodes is not available when you use the Input qualifier.
- While the collection and presentation feature is active,
all on-screen menu options continue to operate as usual.
This includes the time-plot, scatter-plot, screen pause,
and various other options.
- There is no way to exclude the current node from statistics
collection. Log in to another node if you want to do this.
- The cluster collection of per-process stall information
automatically detects the binding or unbinding of processes
to cluster databases. There is no need to manually refresh
the database information on the current node.
- If the database is not currently open on the specified
node, Oracle RMU still attempts to collect cluster
statistics. However, you must open the remote database
prior to regular process attaches.
- When you display any of the per-process screens that
support cluster statistics collection, such as the Stall
Messages screen, you can zoom in on any of the displayed
processes to show which node that process is using.
- Using the Cluster Statistics submenu from the Tools menu,
it is also possible to collect statistics from all open
database nodes using the Collect From Open Database Nodes
menu option. This option simplifies the DBA's job of
remembering where the database is currently open. However,
subsequently opened nodes are not automatically added to
the collection; these must be manually added.
- The cluster statistics collection is an intracluster
feature in that it works only on the same database, using
the same device and directory specification used to run the
initial RMU Show Statistics command (that is, on a shared
disk). The cluster statistics collection does not work
across clusters (intercluster).
- When you replay a binary output file, the screen header
region accurately reflects the number of cluster nodes
whose statistics are represented in the output file.
6 – Examples
Example 1
The following example directs the results of the RMU Show
Statistics command to an output file:
$ RMU/SHOW STATISTICS MF_PERSONNEL/OUTPUT=PERS.LOG
Example 2
The following example formats the binary results created in the
previous example and produces a readable display:
$ RMU/SHOW STATISTICS/INPUT=PERS.LOG
Example 3
The following DCL script shows a complete example of how to
create an excessive stall notification server using the operator
notification facility. To execute this script, submit it to any
queue on the node from which you want to run the script. Supply
the parameters as follows:
o P1 is the database pathname.
o P2 is the completion time.
o P3 is the set of operators to be notified. You must enclose
the list of operators in quotes.
$ VERIFY = F$VERIFY(0)
$ SET NOON
$!
$! Get the database name.
$!
$ IF P1 .EQS. "" THEN INQUIRE P1 "_database"
$!
$! Get the termination date/time.
$!
$ IF P2 .EQS. "" THEN INQUIRE P2 "_until"
$!
$! Get the operator classes.
$!
$ IF P3 .EQS. "" THEN INQUIRE P3 "_operators"
$!
$ RMU/SHOW STATISTICS/TIME=1/NOBROADCAST -
/NOINTERACTIVE /UNTIL="''P2'" /ALARM=5 /NOTIFY='P3 -
'P1
$ VERIFY = F$VERIFY(VERIFY)
$ EXIT
Example 4
You can use the Lock_Timeout or Deadlock qualifiers to construct
a Lock Event Logging server. The following OpenVMS DCL script
shows how to create a server that logs both lock timeout and
lock deadlock events on the MF_PERSONNEL database for the next 15
minutes:
$ RMU/SHOW STATISTICS /NOHISTOGRAM /TIME=1 /NOINTERACTIVE -
_$ /LOCK_TIMEOUT_LOG=TIMEOUT.LOG /DEADLOCK_LOG=DEADLOCK.LOG -
_$ /NOBROADCAST /UNTIL="+15:00" MF_PERSONNEL
Example 5
The following example shows stall log information first with and
then without the lock information:
$ RMU /SHOW STATISTICS /NOINTERACTIVE /STALL_LOG=SYS$OUTPUT: -
_$ DUA0:[DB]MFP.RDB
Oracle Rdb X7.1-00 Performance Monitor Stall Log
Database DPA500:[RDB_RANDOM.RDB_RANDOM_TST_247]RNDDB.RDB;1
Stall Log created 4-SEP-2001 11:27:03.96
11:27:03.96 0002B8A1:1 11:27:03.67 waiting for record 118:2:2 (PR)
State... Process.ID Process.name... Lock.ID. Rq Gr Queue "record 118:2:2"
Blocker: 000220A7 RND_TST_24716 0F019E52 EX Grant
Waiting: 0002B8A1 RND_TST_24715 4500C313 PR Wait
11:27:03.96 0002B8A8:1 11:27:02.32 waiting for record 101:3:0 (EX)
State... Process.ID Process.name... Lock.ID. Rq Gr Queue "record 101:3:0"
Blocker: 000220AD RND_TST_24710 0B00176A PR Grant
Blocker: 000220A7 RND_TST_24716 52018A3F PR Grant
Waiting: 0002B8A8 RND_TST_2474 3C00B5AF EX PR Cnvrt
11:27:03.96 0002B89C:1 11:27:00.15 waiting for record 114:4:1 (PR)
State... Process.ID Process.name... Lock.ID. Rq Gr Queue "record 114:4:1"
Blocker: 000220A7 RND_TST_24716 180033CC EX Grant
Waiting: 0002B89C RND_TST_2479 110066BA PR Wait
$ RMU /SHOW STATISTICS /NOINTERACTIVE /STALL_LOG=SYS$OUTPUT: -
_$ DUA0:[DB]MFP.RDB /OPTIONS=NOLOG_STALL_LOCK
Oracle Rdb X7.1-00 Performance Monitor Stall Log
Database DPA500:[RDB_RANDOM.RDB_RANDOM_TST_247]RNDDB.RDB;1
Stall Log created 4-SEP-2001 11:28:34.68
11:28:34.69 0002B8B8:1 11:28:33.69 waiting for logical area 146 (PR)
11:28:34.69 0002B8A8:1 11:28:32.76 waiting for record 114:4:2 (PR)
11:28:34.69 0002B8B3:1 11:28:33.06 waiting for record 114:4:2 (PR)
11:28:34.69 0002B8B0:1 11:28:31.96 waiting for record 111:7:7 (EX)