Automatically Restarting Software Processes

The LSMS Automatic Software Recovery feature, available as a standard feature for LSMS Release 2.0 and later, detects failures in certain LSMS processes and attempts to restart the processes without the need for manual intervention by the customer. This feature is implemented by the sentryd utility.

Detecting Failure Conditions

Table 1 shows which processes are checked by sentryd and the error conditions for which they are checked.

Processes Monitored by the Automatic Software Recovery Feature
Process	Unintentional Exit	Inability to Perform Defined Tasks	Failed to Initialize During Startup	See section:
EAGLE agents	X	X	X	Automatically Monitoring and Restarting EAGLE Agent Processes
Regional NPAC agents	X	X	X	Automatically Monitoring and Restarting NPAC Agent Processes
OSI	X			Automatically Monitoring and Restarting OSI Process
Service Assurance	X			Automatically Monitoring and Restarting the Service Assurance Process
Local Services Manager	X	X	X	Automatically Monitoring and Restarting Other Processes
Local Data Manager	X	X	X	Automatically Monitoring and Restarting Other Processes
Logger Server	X		X	Automatically Monitoring and Restarting Other Processes
LSMS SNMP Agent	X		X	Automatically Monitoring and Restarting Other Processes
Apache web server	X		X	Automatically Monitoring and Restarting Other Processes
RMTP Manager	X		X	Automatically Monitoring and Restarting the rmtpmgr Process
RMTP Agent	X		X	Automatically Monitoring and Restarting the rmtpagent Process
Report Manager	X		X	Automatically Monitoring and Restarting Other Processes

The sentryd process uses either of the following methods to detect failures:

Verifying that the process has updated its timestamp in the supplemental database periodically
Using standard Linux commands to determine whether a process is running

For more information about specific methods used to detect failures, see the section shown in the last column of Table 5-1.

Reporting Failures Through the Surveillance Feature

If the Surveillance feature is not enabled, sentryd still detects failures and attempts to restart processes, but important information concerning the state of the LSMS is neither displayed nor logged.

To obtain the full benefit of this feature, the Surveillance feature must be enabled. The Surveillance feature displays and logs (in /var/TKLC/lsms/logs/survlog.log) the following notifications regarding the following conditions:

Software failures
Successful recovery of the software
Unsuccessful recovery of the software

Also, whether or not the Surveillance feature is enabled, surveillance agents will restart the sentryd process if it exits abnormally.

Automatically Restarting Processes Hierarchically

Figure 5-1 shows how sentryd restarts processes in a hierarchical order.

Order of Automatically Restarting Processes

This figure illustrates:

Which processes sentryd monitors.
When a failure is detected in a process, sentryd attempts to restart the failed process and all processes shown below it.
The optional Service Assurance process is monitored for failure, but is not restarted by sentryd. Also, if sentryd restarts the OSI process, it stops the Service Assurance process. (The Surveillance feature restarts the Service Assurance process whenever it detects that the Service Assurance process has stopped.)

All recovery procedures start within 60 seconds of failure detection.