Watchdog Service
Service
Provides an API for monitoring critical applications and services for deadlocks and other similar faults. The watchdog for a process is started by calling le_wdog_Kick or le_wdog_Timeout. Once started, the watchdog must be kicked periodically by calling le_wdog_Kick. If the watchdog is not kicked (because it is deadlocked, for example), the watchdog service will attempt to recover the service by killing the process and executing the action specified in the process' watchdogAction (if specified) otherwise its faultAction.
Generally a service which uses the watchdog should set watchdogTimeout to give the default timeout, and watchdogAction to give a recovery action.
Critical services should also set maxWatchdogTimeout. In this case the watchdog will be started on system boot, and cannot be stopped or set longer than the timeout given in maxWatchdogTimeout. This ensures the service is always running as long as the system is running.
- Note
- If maxWatchdogTimeout is not set, no more action is taken if performing the process'
watchdogActiondoesn't recover the process. IfmaxWatchdogTimeoutis specified the system will be rebooted if the process does not recover.
Additionally the watchdog service can be configured to call a callback periodically if the watchdog service process is functioning; i.e. all watchdogs have been kicked and/or non-functioning processes are being recovered. Typically this callback will kick an external watchdog such as /dev/watchdog.
Copyright (C) Sierra Wireless Inc.
A Supervisor API used by the Watchdog daemon to let the Supervisor know that a watched process has timed out
Copyright (C) Sierra Wireless Inc.