add support for MCLAG (sonic-net#453)

* add support for MCLAG

Signed-off-by: shine.chen <[email protected]>

* add warm-reboot support for ICCPd

Signed-off-by: shine.chen <[email protected]>

* ensure iccpd is there before stop iccpd

Signed-off-by: shine.chen <[email protected]>

* fix service function

* remove unused comment

* refactor code according to feature management mechanism

Signed-off-by: shine.chen <[email protected]>

Author: shine4chen

config/main.py

@@ -409,6 +409,14 @@ def _abort_if_false(ctx, param, value):
     if not value:
         ctx.abort()
 
+def _get_optional_services():
+    config_db = ConfigDBConnector()
+    config_db.connect()
+    optional_services_dict = config_db.get_table('FEATURE')
+    if not optional_services_dict:
+        return None
+    return optional_services_dict.keys()
+
 def _stop_services():
     # on Mellanox platform pmon is stopped by syncd
     services_to_stop = [
@@ -439,6 +447,17 @@ def _stop_services():
             log_error("Stopping {} failed with error {}".format(service, e))
             raise
 
+    # For optional services they don't start by default
+    for service in _get_optional_services():
+        (out, err) = run_command("systemctl status {}".format(service), return_output = True)
+        if not err and 'Active: active (running)' in out:
+            try:
+                click.echo("Stopping service {} ...".format(service))
+                run_command("systemctl stop {}".format(service))
+            except SystemExit as e:
+                log_error("Stopping {} failed with error {}".format(service, e))
+                raise
+
 def _reset_failed_services():
     services_to_reset = [
         'bgp',
@@ -474,6 +493,17 @@ def _reset_failed_services():
             log_error("Failed to reset failed status for service {}".format(service))
             raise
 
+    # For optional services they don't start by default
+    for service in _get_optional_services():
+        (out, err) = run_command("systemctl is-enabled {}".format(service), return_output = True)
+        if not err and 'enabled' in out:
+            try:
+                click.echo("Resetting failed status for service {} ...".format(service))
+                run_command("systemctl reset-failed {}".format(service))
+            except SystemExit as e:
+                log_error("Failed to reset failed status for service {}".format(service))
+                raise
+
 def _restart_services():
     # on Mellanox platform pmon is started by syncd
     services_to_restart = [
@@ -508,6 +538,17 @@ def _restart_services():
             log_error("Restart {} failed with error {}".format(service, e))
             raise
 
+    # For optional services they don't start by default
+    for service in _get_optional_services():
+        (out, err) = run_command("systemctl is-enabled {}".format(service), return_output = True)
+        if not err and 'enabled' in out:
+            try:
+                click.echo("Restarting service {} ...".format(service))
+                run_command("systemctl restart {}".format(service))
+            except SystemExit as e:
+                log_error("Restart {} failed with error {}".format(service, e))
+                raise
+
 def is_ipaddress(val):
     """ Validate if an entry is a valid IP """
     if not val:

scripts/fast-reboot

@@ -430,6 +430,24 @@ debug "Stopped  bgp ..."
 docker kill lldp &> /dev/null || debug "Docker lldp is not running ($?) ..."
 systemctl stop lldp
 
+if [[ "$REBOOT_TYPE" = "fast-reboot" ]]; then
+    if echo $(docker ps) | grep -q iccpd; then
+        docker kill iccpd > /dev/null || [ $? == 1 ]
+    fi
+fi
+
+# Stop iccpd gracefully
+if [[ "$REBOOT_TYPE" = "warm-reboot" || "$REBOOT_TYPE" = "fastfast-reboot" ]]; then
+    if echo $(docker ps) | grep -q iccpd; then
+        debug "Stopping iccpd ..."
+        # Send USR1 signal to iccpd to stop it
+        # It will prepare iccpd for warm-reboot
+        # Note: We must send USR1 signal before syncd, or some state of iccpd maybe lost
+        docker exec -i iccpd pkill -USR1 iccpd || [ $? == 1 ] > /dev/null
+        debug "Stopped  iccpd ..."
+    fi
+fi
+
 if [[ "$REBOOT_TYPE" = "fast-reboot" ]]; then
     # Kill teamd processes inside of teamd container with SIGUSR2 to allow them to send last LACP frames
     # We call `docker kill teamd` to ensure the container stops as quickly as possible,