 |
HP OpenVMS System Management Utilities Reference
Manual
For playback requests, a collection event occurs each time a new
interval is encountered in the input file of previously recorded data.
A recording event (if requested) does not necessarily follow
immediately as it does in live collection. Its frequency is still
governed by the /INTERVAL qualifier; the specified /INTERVAL value is
interpreted in terms of the /INTERVAL value specified when the input
file was created. The new value must be an integral multiple of the
original value. A recording event is then triggered every time an
interval is encountered in the input file that is the appropriate
multiple of the original interval.
For playback requests, occurrences of display events (if requested) are
indicated in exactly the same way as recording events (with the
/INTERVAL qualifier) and immediately follow recording events (if both
are specified). The actual length of time a displayed image remains on
the screen is still specified with the /VIEWING_TIME qualifier, but,
unlike the live collection case, this qualifier is not used to signal a
display event.
The following table summarizes which qualifiers cause the various
MONITOR events:
| Event |
Live Collection Qualifier |
Playback Qualifier |
|
Collection
|
/INTERVAL
|
Original /INTERVAL value (from file)
|
|
Recording
|
/INTERVAL
|
/INTERVAL
|
|
Display
|
/VIEWING_TIME
|
/INTERVAL
|
Note that, for live requests, the collection interval is defined as the
number of seconds from the end of one collection event to the beginning
of the next. A collection event includes collection for all requested
classes on all nodes specified. (For multiple-node requests, a
collection event must complete on all nodes before a new event is
initiated.) Therefore, the elapsed time from the beginning of one
collection event to the beginning of the next is the interval value
plus the time it takes to do the collection. For some requests, notably
those including many classes or the PROCESSES, RMS, CLUSTER, or SYSTEM
classes, collection time can be significant.
For /INPUT requests, the interval value defaults to the value specified
in the input recording file. The default for monitoring the running
system is 3 seconds for all classes except ALL_CLASSES, CLUSTER, and
SYSTEM, which have a default of 6 seconds.
/NODE=(nodename,...)
Specifies the nodes (up to 48 in a cluster) for which data is to be
collected. If you specify more than one name, separate the names with
commas, and enclose the list in parentheses.
Remote monitoring in an OpenVMS Cluster environment might not be
compatible for nodes that are running different OpenVMS versions. The
following table shows the compatibility of versions for remote
monitoring:
| |
OpenVMS Alpha and VAX Version 6.0 and later |
OpenVMS Alpha Version 1.5 and VAX Version 5.n |
|
OpenVMS Alpha and VAX Version 6.0 or later
|
Yes
|
No
|
|
OpenVMS Alpha Version 1.5 and VAX Version 5.
n
|
No
|
Yes
|
To obtain data from an incompatible remote node, record the data on the
remote node and then use the MONITOR playback feature to examine the
data on the local node. The HP OpenVMS System Manager's Manual describes remote monitoring.
If you specify multiple node names with multiple system classes,
MONITOR displays one class at a time for each node. For example, the
command MONITOR/NODE=(NODE_A,NODE_B) STATES,MODES generates STATES data
for NODE_A and NODE_B and then MODES data.
/OUTPUT=file-spec
Used with the CONVERT command, this qualifier specifies the name of the
converted recording file. The default specification is MONITOR.DAT.
File lists are not permitted.
Recording files produced using MONITOR prior to VMS Version 5.0 must be
converted to the current format before they can be played back by the
current MONITOR version.
/RECORD[=file-spec]
/NORECORD (default)
Specifies that a binary disk file be created containing all collected
data for the request. Note that recording is restricted to files on
disks. No wildcard characters are allowed in the file specification. If
you omit the file type, the default file type is .DAT. If you omit the
file specification, output is generated to a file named MONITOR.DAT in
the current default device and directory. If you specify an existing
file but omit the version number, a new version of the file is created.
The output consists of all data for the requested classes, regardless
of the classname qualifiers specified. Note that recording file output
is not produced when a multifile summary is requested.
/SUMMARY[=file-spec]
/NOSUMMARY (default)
Specifies that an ASCII disk file be created containing summary
statistics on all data collected for this request. If the optional file
specification is omitted, it defaults to MONITOR.SUM.
The summary file, generated at the end of monitoring, contains one or
more pages of output for each requested class. The format of each page
is similar to that of display output and is determined by the classname
qualifiers. The /ALL qualifier is applied to all class names for which
no other qualifier is specified.
/VIEWING_TIME=seconds
Specifies the duration for each screen image display for /DISPLAY
requests. Values can range from 1 to 9,999,999.
If you are monitoring the running system, /VIEWING_TIME defaults to the
/INTERVAL value. If you specify /INPUT, and you are monitoring a
recording file, /VIEWING_TIME defaults to 3 seconds.
Effective viewing time varies, however, depending on whether you are
running MONITOR on your local system or on a remote node.
(Remote in this context refers to the use of the SET HOST
command to access another node.) For remote access, the time required
to display the screen is included in the viewing time, while for local
access, this time is not included. Therefore, use a larger viewing time
than the 3-second default when running MONITOR on a remote system. The
value appropriate for remote access depends on your terminal baud rate.
For a 9600--baud terminal line, 6 seconds is a reasonable viewing time.
Note also that the time between full screens of data for the PROCESSES
display is controlled by this qualifier.
MONITOR ALL_CLASSES
The MONITOR ALL_CLASSES command initiates monitoring of statistics for
all classes except the CLUSTER and RMS classes.
Format
MONITOR ALL_CLASSES
Command Qualifiers
/qualifier[,...]
One or more qualifiers as described in the Command Qualifier
Descriptions section.
Classname Qualifiers
/ALL
Specifies that a table of all available statistics (current, average,
minimum, and maximum) is to be included in the display and summary
output. For summary output, this qualifier is the default for all
classes; otherwise, it is the default for all classes except CLUSTER,
MODES, PROCESSES, STATES, SYSTEM, and VECTOR.
/AVERAGE
Specifies that a bar graph of average statistics is to be included in
the display and summary outputs.
/CURRENT
Specifies that a bar graph of current statistics is to be included in
the display and summary outputs. The /CURRENT qualifier is the default
for the CLUSTER, MODES, STATES, SYSTEM, and VECTOR classes.
/MAXIMUM
Specifies that a bar graph of maximum statistics is to be included in
the display and summary outputs.
/MINIMUM
Specifies that a bar graph of minimum statistics is to be included in
the display and summary outputs.
Description
If you do not specify any qualifiers with the ALL_CLASSES parameter,
normal default output is produced for each class. The qualifiers have
no effect on display of the PROCESSES class.
Note that the default interval is 6 seconds.
The MONITOR ALL_CLASSES command is particularly useful for playback of
recording files because it eliminates the need to specify the
particular classes of performance data the recording file contains. To
override any of the default qualifiers, specify the class name with the
qualifier after specifying ALL_CLASSES.
Example
|
MONITOR> MONITOR/INPUT=SYS$MANAGER:LOADBAL.DAT ALL_CLASSES,PROCESSES/TOPCPU
|
This command initiates playback of the recording file
SYS$MANAGER:LOADBAL.DAT. All data contained in the file will be
displayed.
MONITOR CLUSTER
The MONITOR CLUSTER command initiates monitoring of the CLUSTER
statistics class, which shows clusterwide CPU, memory, disk, and
locking activity.
Format
MONITOR CLUSTER
Command Qualifiers
/qualifier[,...]
One or more qualifiers as described in the Command Qualifier
Descriptions section.
Classname Qualifiers
/ALL
Specifies that a table of all available statistics (current, average,
minimum, and maximum) is to be included in the display and summary
output. For summary output, this qualifier is the default for all
classes; otherwise, it is the default for all classes except CLUSTER,
MODES, PROCESSES, STATES, SYSTEM, and VECTOR.
/AVERAGE
Specifies that a bar graph of average statistics is to be included in
the display and summary outputs.
/CURRENT
Specifies that a bar graph of current statistics is to be included in
the display and summary outputs. The /CURRENT qualifier is the default
for the CLUSTER, MODES, STATES, SYSTEM, and VECTOR classes.
/MAXIMUM
Specifies that a bar graph of maximum statistics is to be included in
the display and summary outputs.
/MINIMUM
Specifies that a bar graph of minimum statistics is to be included in
the display and summary outputs.
Description
For the CLUSTER class, MONITOR collects data items for up to 48 nodes
in a cluster. Because this class combines the most significant
clusterwide performance statistics in a single display, it is
particularly useful to cluster managers and other users seeking an
overview of cluster activity.
MONITOR does not recognize nodes that enter the cluster while a request
is active. MONITOR, therefore, does not collect data for these nodes.
You cannot specify the CLUSTER class in the same request with any other
class.
In a multifile summary request, the classes CLUSTER and PROCESSES are
ignored. If these classes are the only classes specified on the command
line, MONITOR does not recognize them and displays a "no classes
specified" error message. MONITOR does not recognize these classes
if they are the only classes specified on the command line, and
displays a "no classes specified" error message.
The CLUSTER class includes the following data items:
| Data Item |
Description |
|
CPU Busy
|
Percentage of CPU in use; includes activity in all processor modes
(except Idle Time) for each node.
|
|
Percent Memory In Use
|
Memory in use on each node; calculated by dividing the Free List Size
by total available memory and subtracting the result from 100%.
|
|
I/O Operation Rate
|
Total rate of disk I/O operations on each disk by all nodes currently
active in the request.
In cluster configurations, the MSCP server software makes locally
attached and HSC disks available to other nodes. A node uses
remote access to a disk when it accesses the disk through
another VAX node (using the MSCP server). A node uses
direct access to a disk when it directly accesses a locally
attached or HSC disk.
An "R" following the device name indicates that the
displayed statistics represent I/O operations requested by nodes using
remote access.
If an "R" does not appear after the device name, the
displayed statistics represent I/O operations issued by nodes with
direct access. These I/O operations might include those issued by the
MSCP server on behalf of remote requests.
|
|
Total ENQ/DEQ Rate
|
Sum of all local, incoming, and outgoing ENQs, DEQs, and conversions.
|
Two display formats are provided, depending on the classname qualifier
specified:
- A tabular style format for the /ALL qualifier
- A bar graph style format for the /AVERAGE, /CURRENT, /MAXIMUM, and
/MINIMUM qualifiers
Beginning in OpenVMS Version 7.3, the range of rate fields has been
increased in the MONITOR CLUSTER screen display as follows:
| Rate Name |
Old Rate |
New Rate |
|
I/O Operation
|
0 - 25 - 50 - 75 - 100
|
0 - 125 - 250 - 375 - 500
|
|
Lock
|
Scale from 0 to 500
|
Scale from 0 to 1000
|
Note to Cluster Managers on MONITOR_SERVER Process
When users enter the MONITOR CLUSTER command, MONITOR activates the
image SYS$SYSTEM:VPM.EXE, which creates a process called
MONITOR_SERVER on each remote cluster node. (If users
specify the /NODE qualifier with the MONITOR CLUSTER command or with
any command of the form MONITOR class name, MONITOR creates the process
only on the specified nodes.) The server process gathers data from
remote nodes for live display or to record on the local node. To ensure
accurate and timely data collection, the process is started at priority
15. Because server processes consume minimal resources, they have no
significant effect on system performance.
By default, MONITOR_SERVER processes are started in the system DECnet
account, which is created when the NETCONFIG.COM command procedure
executes at bootstrap time. If this account is not present on your
system, you must either create it by executing NETCONFIG.COM, or
specify another account in which the server processes can be started.
If you want to start the processes in another account, use the
following sequence of commands to define VPM as known object 51 in the
DECnet database and associate the object with the desired account:
$ SET PROCESS/PRIVILEGE=SYSPRV
$ RUN SYS$SYSTEM:NCP
NCP> DEFINE OBJECT VPM NUMBER 51 -
_ FILE SYS$SYSTEM:VPM.EXE -
_ PROXY NONE -
_ ACCOUNT account -
_ USER user-id -
_ PASSWORD password
NCP> SET OBJECT VPM NUMBER 51 -
_ FILE SYS$SYSTEM:VPM.EXE -
_ PROXY NONE -
_ ACCOUNT account -
_ USERNAME user-id -
_ PASSWORD password
NCP> EXIT
$ SET PROCESS/PRIVILEGE=NOSYSPRV
|
For each server process, MONITOR creates a log file on the local node
to which information about server connection activity, including error
messages, is written. Note that error messages are written to the file
only when errors occur. A single version is maintained for the life of
the system. The default file specification has the form
SYS$COMMON:[SYSMGR]VPM$nodename.LOG. The node name portion of the
specification identifies the node on which the MONITOR_SERVER process
has been started.
If you want to change the default specification, you can redefine the
executive-mode logical name VPM$LOG_FILE in the system logical name
table on the appropriate nodes. For example, if you wanted to write
server error logging data to the file WRKD:[MONSERVER]VPM_ERRORS.LOG,
you would define VPM$LOG_FILE as follows:
$ DEFINE/SYSTEM/EXECUTIVE_MODE VPM$LOG_FILE -
_$ WRKD:[MONSERVER]VPM_ERRORS.LOG
|
To direct to a single file data for all MONITOR_SERVER processes on the
cluster, you could assign the logical name the same value on each
member system. Note that because the log files are created as shared
sequential files, multiple server processes can access them
simultaneously.
If you routinely monitor your cluster, you can reduce server startup
time significantly by creating MONITOR_SERVER processes on each member
node at bootstrap time and maintaining the processes for the life of
the system. To do so, add the following lines to the appropriate
site-independent startup command files:
$ DEFINE/SYSTEM/EXECUTIVE_MODE VPM$SERVER_LIVE TRUE
$ RUN/DETACH/PAGE_FILE=10000 SYS$SYSTEM:VPM.EXE
|
You can enter these commands interactively at any time if you have the
following privileges: ALTPRI, NETMBX, PSWAPM, SYSNAM, SYSPRV, and
TMPMBX.
Example
|
MONITOR> MONITOR CLUSTER/ALL
OpenVMS Monitor Utility
CLUSTER STATISTICS
on node CURLEY
29-APR-2003 12:25:13
CPU Busy CUR AVE MIN MAX
LARRY 100.00 100.00 100.00 100.00
CURLEY 100.00 99.83 100.00 100.00
MOE 8.52 8.50 8.52 8.52
OpenVMS Monitor Utility
CLUSTER STATISTICS
on node CURLEY
29-APR-2003 12:25:19
%Memory In Use CUR AVE MIN MAX
MOE 88.00 88.00 88.00 88.00
LARRY 78.00 78.00 77.00 78.00
CURLEY 72.00 72.50 72.00 72.00
OpenVMS Monitor Utility
CLUSTER STATISTICS
on node CURLEY
29-APR-2003 12:25:25
I/O Operation Rate CUR AVE MIN MAX
$111$DUA7: (DECEIT) SQMCLUSTERV4 0.48 6.53 0.48 10.41
$111$DUA6: (DECEIT) QUALD 1.93 1.07 0.00 1.93
$111$DUA4: (DECEIT) PAGESWAPDISK 1.44 0.96 0.00 1.44
$111$DUA2: (DECEIT) TSDPERF 0.32 0.53 0.16 1.12
LARRY$DRA3: QUALQUEST 0.00 0.21 0.00 0.64
MOE$DMA1: UVMSQAR 0.00 0.00 0.00 0.00
MOE$DRA5: USER01 0.00 0.00 0.00 0.00
LARRY$DRA4: TIMEDEV 0.00 0.00 0.00 0.00
LARRY$DBB3: REGLIB 0.00 0.00 0.00 0.00
$111$DUA3: (DECEIT) DUMPDISK 0.00 0.00 0.00 0.00
$111$DUA5: (DECEIT) BPMDISK 0.00 0.00 0.00 0.00
$111$DJA8: (DECEIT) ORLEAN 0.00 0.00 0.00 0.00
$111$DJA10: (DECEIT) QMISDATABASE 0.00 0.00 0.00 0.00
$111$DJA9: (DECEIT) MPI$DATA 0.00 0.00 0.00 0.00
|
OpenVMS Monitor Utility
CLUSTER STATISTICS
on node CURLEY
29-APR-2003 12:25:56
Tot ENQ/DEQ Rate CUR AVE MIN MAX
MOE 7.90 14.92 0.00 43.12
LARRY 20.48 14.64 0.00 46.92
CURLEY 1.93 13.29 0.00 57.30
|
|
The preceding example shows the tabular style format for the CLUSTER
display.
|
MONITOR> MONITOR CLUSTER/CURRENT
Statistic: CURRENT OpenVMS Monitor Utility 5-JUN-2003
10:46:53
CLUSTER STATISTICS
CPU | MEMORY
|
CPU Busy 0 25 50 75 100|%Memory In Use 0 25 50 75 100
+----+----+----+----+| +----+----+----+----+
BRS004 100 |********************|BRS004 37 |*******
| | |
| | |
| | |
| | |
| | |
-----------------------------------------+--------------------------------------
DISK | LOCK
|
I/O Operation Rate 0 125 250 375 500|Tot ENQ/DEQ Rate 0 250 500 750 1000
+----+----+----+----+| +----+----+----+----+
$1$DIA1: 52 |** |BRS004 183 |***
| | |
| | |
| | |
| | |
| | |
|
|
The preceding example shows the bar graph style format for a
CLUSTER/CURRENT display.
MONITOR DECNET
The MONITOR DECNET command initiates monitoring of the DECNET class,
which includes information about DECnet for OpenVMS network activity.
Format
MONITOR DECNET
Command Qualifiers
/qualifier[,...]
One or more qualifiers as described in the Command Qualifier
Descriptions section.
Classname Qualifiers
/ALL
Specifies that a table of all available statistics (current, average,
minimum, and maximum) is to be included in the display and summary
output. For summary output, this qualifier is the default for all
classes; otherwise, it is the default for all classes except CLUSTER,
MODES, PROCESSES, STATES, SYSTEM, and VECTOR.
/AVERAGE
Specifies that a bar graph of average statistics is to be included in
the display and summary outputs.
/CURRENT
Specifies that a bar graph of current statistics is to be included in
the display and summary outputs. The /CURRENT qualifier is the default
for the CLUSTER, MODES, STATES, SYSTEM, and VECTOR classes.
/MAXIMUM
Specifies that a bar graph of maximum statistics is to be included in
the display and summary outputs.
/MINIMUM
Specifies that a bar graph of minimum statistics is to be included in
the display and summary outputs.
Description
The DECNET class consists of the following data items:
| Data Item |
Description |
|
Arriving Local Packet Rate
|
Rate at which local packets are being received.
|
|
Departing Local Packet Rate
|
Rate at which local packets are being sent.
|
|
Arriving Transit Packet Rate
|
Rate at which transit packets are arriving.
|
|
Transit Congestion Loss Rate
|
Rate of transit congestion loss.
|
|
Receiver Buffer Failure Rate
|
Rate of receiver buffer failures.
|
Example
|
MONITOR> MONITOR DECNET
OpenVMS Monitor Utility
DECNET STATISTICS
on node SAMPLE
29-APR-2003 22:22:44
CUR AVE MIN MAX
Arriving Local Packet Rate 9.54 5.08 0.00 11.25
Departing Local Packet Rate 9.22 4.66 0.00 10.92
Arriving Trans Packet Rate 0.00 0.00 0.00 0.00
Trans Congestion Loss Rate 0.00 0.00 0.00 0.00
Receiver Buff Failure Rate 0.00 0.00 0.00 0.00
|
|
This example shows that arriving and departing network packet rates
(including control packets) are roughly equivalent, and that network
activity is currently at a level higher than the average since
monitoring began, but not at its highest point.
|
