watchdogd v3.2-rc1
Pre-release
Pre-release
Changes
- Issue #17: When the process supervisor is enabled
watchdogd
now always runs with elevated RT priority. Previous releases changed toSCHED_RR
only when the first supervised process connected, and conversely disbled RT prio when the last process disconnected. This change gives a more predictable behavior and also meanswatchdogd
can be relied upon until the system has been properly diagnosed - If the (optional) supervisor script returns OK (0) the timer for the offending process is now disarmed and the system is not rebooted.
- Retry handover from Finit buit-in watchdog if first attempt fails
- New generic script monitor, thanks to Tom Deblauwe. Can periodically call a site specific script, with timeout in case the script hangs
Fixes
- Fix #16: Only force reboot on exit if
watchdogd
is enabled - When disabling and the re-enabling
watchdogd
using the API the daemon was sometimes stopped by Finit. This happened because the daemon re-issued a watchdog handover signal to Finit. The fix is to only do the handover once. - When re-enabling
watchdogd
the supervisor was not properly elevating the RT priority, instead it remained as aSCHED_OTHER
process. This fix makes sure to save and re-use the configured RT priority.