2.8.1 Monitoring processes and services
- Organization of this subsection
(1) Monitorable processes and services
APM can monitor services whose name is 100 bytes or less long. The maximum length of the names of monitorable processes varies depending on the OS. The following table lists (by OS) the maximum length of the names of monitorable processes.
OS |
Executable file name |
Command line name |
---|---|---|
HP-UX (IPF) |
14 bytes |
1,020 bytes |
Solaris |
15 bytes |
79 bytes |
AIX |
256 bytes |
2,016 bytes |
Linux |
15 bytes |
2,017 bytes |
Windows |
256 bytes |
256 bytes |
The values in Table 2-26 indicate the maximum length of a process name that APM can recognize. For example, in a system running HP-UX (IPF) OS, a process whose executable file whose name is 20 bytes long is recognized by the first 14 bytes of its name. Therefore, if you want to monitor a process by its executable file name, specify a name that is no more than 14 bytes long or a name that includes a wildcard. If a monitoring process name contains a wildcard, APM monitors any processes that match the specified name up to the recognizable length.
To check the process names recognized by APM, execute the apmproclist command. For details about the apmproclist command, see apmproclist in 5. Commands.
-
Process monitoring in Windows
APM for Windows monitors processes by obtaining Windows performance data.
(2) Events issued by APM
APM issues events to SSO when the following incidents occur:
-
Starting or stopping of APM
-
Status change of a process or service (when a process starts or stops or when the operating status of a service changes)
-
Termination of status check of a process or service
-
Change of the monitoring interval
-
Starting or stopping of monitoring
(3) Event delay
APM uses the UDP protocol to issue events. However, if APM is monitoring processes periodically, events may be issued in the wrong order or even lost. Particularly if APM issues several events to the same server in succession, it is more likely that events will be lost or shuffled. APM can therefore delay the issuance of the second and subsequent events by the specified number of seconds. Using this facility enables you to delay the process status events identified by the IDs 1000005 to 1000011 and the service status events identified by the IDs 1000016 to 1000023. The following figure shows overview of event delay.
|
Set the event delay interval in the event-delay configuration file. For details, see 6.4.5 Event-delay configuration file (apmdelay.conf). When events are reported at the TCP, however, the event delay facility is invalid.
(4) TCP reporting of events
If any event is lost, you can select the TCP protocol to be used for event notification.
(a) How to specify TCP reporting of events
-
Settings in APM
To use UDP for event notification from APM to SSO, disable the TCPSMODE: key in the event TCP notification definition file (apmtcpsend.conf). To do this, set the TCPSMODE: key to OFF.
For details about the event TCP notification definition file, see 6.4.7 Event TCP notification definition file (apmtcpsend.conf).