sonic-buildimage/files/image_config/monit/conf.d/sonic-host

###############################################################################
## Monit configuration for SONiC host OS
##
## This includes system-level monitoring as well as processes which
## run in the host OS (i.e., not inside a Docker container)
###############################################################################

check filesystem root-overlay with path /
    if space usage > 90% for 10 times within 20 cycles then alert repeat every 1 cycles

check filesystem var-log with path /var/log
    if space usage > 90% for 10 times within 20 cycles then alert repeat every 1 cycles

check system $HOST
    if memory usage > 90% for 10 times within 20 cycles then alert repeat every 1 cycles
    if cpu usage (user) > 90% for 10 times within 20 cycles then alert repeat every 1 cycles
    if cpu usage (system) > 90% for 10 times within 20 cycles then alert repeat every 1 cycles

check process rsyslog with pidfile /var/run/rsyslogd.pid
    start program = "/bin/systemctl start rsyslog.service"
    stop program = "/bin/systemctl stop rsyslog.service"
    if totalmem > 800 MB for 10 times within 20 cycles then restart

# route_check.py Verify routes between APPL-DB & ASIC-DB are in sync.
# For any discrepancy, details are logged and a non-zero code is returned
# which would trigger a monit alert.
# Hence for any discrepancy, there will be log messages for "ERR" level
# from both route_check.py & monit.
#
check program routeCheck with path "/usr/local/bin/route_check.py"
    every 5 cycles
    if status != 0 for 3 cycle then alert repeat every 1 cycles

# dualtor_neighbor_check.py: script to check if the neighbor entries in APPL_DB
# has correct neighbor or tunnel route entries in ASIC_DB based on the mux
# states.
check program dualtorNeighborCheck with path "/usr/local/bin/dualtor_neighbor_check.py -o SYSLOG -s ERROR"
    every 5 cycles
    if status != 0 for 3 cycle then alert repeat every 1 cycles

# Check if /etc & /home are writable. If not, make them writable.
# Raise syslog error message, in case of underlying issues
#
check program diskCheck with path "/usr/local/bin/disk_check.py"
    every 5 cycles
    if status != 0 for 3 cycle then alert repeat every 1 cycles

check program container_checker with path "/usr/bin/container_checker"
    if status != 0 for 5 times within 5 cycles then alert repeat every 1 cycles

# vnet_route_check.py: tool that verifies VNET routes consistancy between SONiC and vendor SDK DBs.
check program vnetRouteCheck with path "/usr/local/bin/vnet_route_check.py"
    every 5 cycles
    if status != 0 for 3 cycle then alert repeat every 1 cycles

# memory_check tool that verifies that memory usage does not cross the threshold or invokes techsupport.
check program memory_check with path "/usr/local/bin/memory_threshold_check.py"
    if status == 2 for 10 times within 20 cycles then exec "/usr/local/bin/memory_threshold_check_handler.py"