Improve health probe logging for nova and neutron

1. Log specific compute services failing rabbitMQ socket tests in nova
   health probe
2. Log specific compute services failing Database socket tests in nova
   health probe
3. Make log level configurable for nova and neutron health probes

Change-Id: I5e5d909d598af734596eb1732ae42808c1f6cd12

neutron/Chart.yaml

@@ -14,7 +14,7 @@ apiVersion: v1
 appVersion: v1.0.0
 description: OpenStack-Helm Neutron
 name: neutron
-version: 0.2.10
+version: 0.2.11
 home: https://docs.openstack.org/neutron/latest/
 icon: https://www.openstack.org/themes/openstack/images/project-mascots/Neutron/OpenStack_Project_Neutron_vertical.png
 sources:

neutron/templates/bin/_health-probe.py.tpl

@@ -53,7 +53,7 @@ rpc_timeout = int(os.getenv('RPC_PROBE_TIMEOUT', '60'))
 rpc_retries = int(os.getenv('RPC_PROBE_RETRIES', '2'))
 rabbit_port = 5672
 tcp_established = "ESTABLISHED"
-log.logging.basicConfig(level=log.ERROR)
+log.logging.basicConfig(level=log.{{ .Values.health_probe.logging.level }})
 
 
 def _get_hostname(use_fqdn):

neutron/values.yaml

@@ -2516,6 +2516,10 @@ network_policy:
 
 helm3_hook: true
 
+health_probe:
+  logging:
+    level: ERROR
+
 manifests:
   certificates: false
   configmap_bin: true

nova/Chart.yaml

@@ -14,7 +14,7 @@ apiVersion: v1
 appVersion: v1.0.0
 description: OpenStack-Helm Nova
 name: nova
-version: 0.2.29
+version: 0.2.30
 home: https://docs.openstack.org/nova/latest/
 icon: https://www.openstack.org/themes/openstack/images/project-mascots/Nova/OpenStack_Project_Nova_vertical.png
 sources:

nova/templates/bin/_health-probe.py.tpl

@@ -161,8 +161,10 @@ def test_tcp_socket(service):
 
     if service in dict_services:
         proc = dict_services[service]
+        transport = oslo_messaging.TransportURL.parse(cfg.CONF)
         if r_ports and tcp_socket_status(proc, r_ports) == 0:
-            sys.stderr.write("RabbitMQ socket not established")
+            sys.stderr.write("RabbitMQ socket not established for service "
+                             "%s with transport %s" % (proc, transport))
             # Do not kill the pod if RabbitMQ is not reachable/down
             if not cfg.CONF.liveness_probe:
                 sys.exit(1)
@@ -170,7 +172,8 @@ def test_tcp_socket(service):
         # let's do the db check
         if service != "compute":
             if d_ports and tcp_socket_status(proc, d_ports) == 0:
-                sys.stderr.write("Database socket not established")
+                sys.stderr.write("Database socket not established for service "
+                                 "%s with transport %s" % (proc, transport))
                 # Do not kill the pod if database is not reachable/down
                 # there could be no socket as well as typically connections
                 # get closed after an idle timeout
@@ -194,7 +197,7 @@ def test_rpc_liveness():
 
     cfg.CONF(sys.argv[1:])
 
-    log.logging.basicConfig(level=log.ERROR)
+    log.logging.basicConfig(level=log.{{ .Values.health_probe.logging.level }})
 
     try:
         transport = oslo_messaging.get_transport(cfg.CONF)

nova/values.yaml

@@ -2556,6 +2556,10 @@ network_policy:
 # set helm3_hook: false when using the helm2 binary.
 helm3_hook: true
 
+health_probe:
+  logging:
+    level: ERROR
+
 manifests:
   certificates: false
   configmap_bin: true

releasenotes/notes/neutron.yaml

@@ -24,4 +24,5 @@ neutron:
   - 0.2.8 Add Victoria and Wallaby releases support
   - 0.2.9 Add option to disable helm.sh/hook annotations
   - 0.2.10 Update htk requirements repo
+  - 0.2.11 Improve health probe logging
 ...

releasenotes/notes/nova.yaml

@@ -50,4 +50,5 @@ nova:
   - 0.2.27 Add tls1.2 minimum version to tls overrides
   - 0.2.28 Move ssl_minimum_version to console section
   - 0.2.29 Remove ssh-config
+  - 0.2.30 Improve health probe logging
 ...