From bb6c4111fdacc75025f7bc6a10db40faa89bd578 Mon Sep 17 00:00:00 2001
From: Chris Friesen <chris.friesen@windriver.com>
Date: Mon, 9 Apr 2018 17:14:40 -0600
Subject: [PATCH] Delay starting compute services until controller services are
 up

On AIO-SX startup there are many services that need to get started up.
We have noted experimentally that there can be significant CPU
congestion if the controller services are still initializing when we
start up the compute services.

With this change we will wait for the "cloud-services" service-group to
be active before starting up the compute services.

This has minimal effect on overall startup times because the controller
services can continue to run on all host CPUs until the cloud services
start up, at which point all host services are restricted to the
platform
CPUs.

Change-Id: Ie9f129fb078e19883cab910b3f411813237d376b
---
 computeconfig/computeconfig/compute_config | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/computeconfig/computeconfig/compute_config b/computeconfig/computeconfig/compute_config
index cdd4a46864..37747d24b9 100644
--- a/computeconfig/computeconfig/compute_config
+++ b/computeconfig/computeconfig/compute_config
@@ -98,12 +98,11 @@ get_ip()
 
 wait_for_controller_services()
 {
-    SERVICE="platform-nfs-ip"
     while [ "$SECONDS" -le "$DELAY_SEC" ]
     do
-        # Check to make sure the service is enabled
-        sm-query service ${SERVICE} | grep -q "enabled-active"
-        if [ $? -eq 0 ]
+        # Check to make sure the cloud-services group is enabled
+        OUTPUT=`sm-query service-group cloud-services`
+        if [ "$OUTPUT" == "cloud-services active" ]
         then
            return 0
         fi