Gangams/add support cri runtime docker env

installer/conf/azm-containers-parser.conf

@@ -0,0 +1,19 @@
+[PARSER]
+    Name         docker
+    Format       json
+    Time_Key     time
+    Time_Format  %Y-%m-%dT%H:%M:%S.%L
+    Time_Keep    On
+    # Command      |  Decoder | Field | Optional Action
+    # =============|==================|=================
+    Decode_Field_As   escaped    log
+
+[PARSER]
+    # http://rubular.com/r/tjUt3Awgg4
+    Name cri
+    Format regex
+    Regex ^(?<time>[^ ]+) (?<stream>stdout|stderr) (?<logtag>[^ ]*) (?<log>.*)$
+    Time_Key    time
+    Time_Format %Y-%m-%dT%H:%M:%S.%L
+    Time_Keep    On   
+

installer/conf/td-agent-bit-rs.conf

@@ -1,7 +1,7 @@
 [SERVICE]
     Flush         30
     Log_Level     info
-    Parsers_File  /etc/td-agent-bit/parsers.conf
+    Parsers_File  /etc/opt/microsoft/docker-cimprov/azm-containers-parser.conf
     Log_File      /var/opt/microsoft/docker-cimprov/log/fluent-bit.log
 
 [INPUT]

installer/conf/td-agent-bit.conf

@@ -2,7 +2,7 @@
     #Default service flush interval is 15 seconds
     ${SERVICE_FLUSH_INTERVAL}
     Log_Level     info
-    Parsers_File  /etc/td-agent-bit/parsers.conf
+    Parsers_File  /etc/opt/microsoft/docker-cimprov/azm-containers-parser.conf
     Log_File      /var/opt/microsoft/docker-cimprov/log/fluent-bit.log
 
 [INPUT]

installer/conf/telegraf.conf

@@ -386,7 +386,7 @@
 #  report_active = true
 #  fieldpass = ["usage_active","cluster","node","host","device"]
 #  taginclude = ["cluster","cpu","node"]
-  
+
 
 
 # Read metrics about disk usage by mount point
@@ -395,7 +395,7 @@
   ## By default stats will be gathered for all mount points.
   ## Set mount_points will restrict the stats to only the specified mount points.
   # mount_points = ["/"]
-  
+
   ## Ignore mount points by filesystem type.
   ignore_fs = ["tmpfs", "devtmpfs", "devfs", "overlay", "aufs", "squashfs"]
   fieldpass = ["free", "used", "used_percent"]
@@ -532,7 +532,8 @@
   name_prefix="container.azm.ms/"
   ## An array of urls to scrape metrics from.
   urls = ["$CADVISOR_METRICS_URL"]
-  fieldpass = ["kubelet_docker_operations", "kubelet_docker_operations_errors"]
+  ## Include "$KUBELET_RUNTIME_OPERATIONS_TOTAL_METRIC", "$KUBELET_RUNTIME_OPERATIONS_ERRORS_TOTAL_METRIC" when we add for support for 1.18
+  fieldpass = ["$KUBELET_RUNTIME_OPERATIONS_METRIC", "$KUBELET_RUNTIME_OPERATIONS_ERRORS_METRIC"]
 
   metric_version = 2
   url_tag = "scrapeUrl"
@@ -578,7 +579,7 @@
   urls = $AZMON_DS_PROM_URLS
 
   fieldpass = $AZMON_DS_PROM_FIELDPASS
-  
+
   fielddrop = $AZMON_DS_PROM_FIELDDROP
 
   metric_version = 2

installer/datafiles/base_container.data

@@ -45,6 +45,8 @@ MAINTAINER:              'Microsoft Corporation'
 /opt/microsoft/omsagent/plugin/DockerApiClient.rb;												source/code/plugin/DockerApiClient.rb;				644; root; root
 /opt/microsoft/omsagent/plugin/DockerApiRestHelper.rb;											source/code/plugin/DockerApiRestHelper.rb;			644; root; root
 /opt/microsoft/omsagent/plugin/in_containerinventory.rb;										source/code/plugin/in_containerinventory.rb;		644; root; root
+/opt/microsoft/omsagent/plugin/kubernetes_container_inventory.rb;							source/code/plugin/kubernetes_container_inventory.rb;		644; root; root
+
 
 /opt/microsoft/omsagent/plugin/out_mdm.rb;                                                  source/code/plugin/out_mdm.rb; 644; root; root
 /opt/microsoft/omsagent/plugin/filter_cadvisor2mdm.rb;                                      source/code/plugin/filter_cadvisor2mdm.rb; 644; root; root
@@ -106,6 +108,7 @@ MAINTAINER:              'Microsoft Corporation'
 /opt/td-agent-bit/bin/out_oms.so;                                intermediate/${{BUILD_CONFIGURATION}}/out_oms.so;  	755; root; root
 /etc/opt/microsoft/docker-cimprov/td-agent-bit.conf;			    installer/conf/td-agent-bit.conf;                    644; root; root
 /etc/opt/microsoft/docker-cimprov/td-agent-bit-rs.conf;			    installer/conf/td-agent-bit-rs.conf;                    644; root; root
+/etc/opt/microsoft/docker-cimprov/azm-containers-parser.conf;	    installer/conf/azm-containers-parser.conf;                    644; root; root
 /etc/opt/microsoft/docker-cimprov/out_oms.conf;			            installer/conf/out_oms.conf;                    644; root; root
 /etc/opt/microsoft/docker-cimprov/telegraf.conf;			        installer/conf/telegraf.conf;                    644; root; root
 /etc/opt/microsoft/docker-cimprov/telegraf-rs.conf;			        installer/conf/telegraf-rs.conf;                    644; root; root

source/code/go/src/plugins/oms.go

@@ -38,6 +38,9 @@ const ResourceIdEnv = "AKS_RESOURCE_ID"
 //env variable which has ResourceName for NON-AKS
 const ResourceNameEnv = "ACS_RESOURCE_NAME"
 
+//env variable which has container run time name
+const ContainerRuntimeEnv = "CONTAINER_RUNTIME"
+
 // Origin prefix for telegraf Metrics (used as prefix for origin field & prefix for azure monitor specific tags and also for custom-metrics telemetry )
 const TelegrafMetricOriginPrefix = "container.azm.ms"
 
@@ -94,7 +97,9 @@ var (
 	//KubeMonAgentEvents skip first flush
 	skipKubeMonEventsFlush bool
 	// enrich container logs (when true this will add the fields - timeofcommand, containername & containerimage)
-	enrichContainerLogs bool
+	enrichContainerLogs bool		
+	// container runtime engine configured on the kubelet
+	containerRuntime string
 )
 
 var (
@@ -750,9 +755,11 @@ func PostDataHelper(tailPluginRecords []map[interface{}]interface{}) int {
 
 		stringMap := make(map[string]string)
 
-		stringMap["LogEntry"] = ToString(record["log"])
+		logEntry := ToString(record["log"])
+		logEntryTimeStamp := ToString(record["time"])			
+		stringMap["LogEntry"] = logEntry
 		stringMap["LogEntrySource"] = logEntrySource
-		stringMap["LogEntryTimeStamp"] = ToString(record["time"])
+		stringMap["LogEntryTimeStamp"] = logEntryTimeStamp
 		stringMap["SourceSystem"] = "Containers"
 		stringMap["Id"] = containerID
 
@@ -797,7 +804,7 @@ func PostDataHelper(tailPluginRecords []map[interface{}]interface{}) int {
 		if dataItem.LogEntryTimeStamp != "" {
 			loggedTime, e := time.Parse(time.RFC3339, dataItem.LogEntryTimeStamp)
 			if e != nil {
-				message := fmt.Sprintf("Error while converting LogEntryTimeStamp for telemetry purposes: %s", e.Error())
+				message := fmt.Sprintf("containerId: %s Error while converting LogEntryTimeStamp for telemetry purposes: %s", dataItem.ID, e.Error())
 				Log(message)
 				SendException(message)
 			} else {
@@ -977,6 +984,11 @@ func InitializePlugin(pluginConfPath string, agentVersion string) {
 		Log("ResourceID=%s", ResourceID)
 		Log("ResourceName=%s", ResourceName)
 	}
+
+	// log runtime info for debug purpose
+	containerRuntime = os.Getenv(ContainerRuntimeEnv)		
+	Log("Container Runtime engine %s", containerRuntime)	
+
 
 	//set useragent to be used by ingestion 
 	docker_cimprov_version := strings.TrimSpace(os.Getenv("DOCKER_CIMPROV_VERSION"))

source/code/plugin/ApplicationInsightsUtility.rb

@@ -6,7 +6,7 @@ class ApplicationInsightsUtility
   require_relative "omslog"
   require_relative "DockerApiClient"
   require_relative "oms_common"
-  require 'yajl/json_gem'
+  require "yajl/json_gem"
   require "base64"
 
   @@HeartBeat = "HeartBeatEvent"
@@ -20,6 +20,7 @@ class ApplicationInsightsUtility
   @@EnvApplicationInsightsKey = "APPLICATIONINSIGHTS_AUTH"
   @@EnvApplicationInsightsEndpoint = "APPLICATIONINSIGHTS_ENDPOINT"
   @@EnvControllerType = "CONTROLLER_TYPE"
+  @@EnvContainerRuntime = "CONTAINER_RUNTIME"
 
   @@CustomProperties = {}
   @@Tc = nil
@@ -73,7 +74,7 @@ def initializeUtility()
           @@Tc = ApplicationInsights::TelemetryClient.new
         elsif !encodedAppInsightsKey.nil?
           decodedAppInsightsKey = Base64.decode64(encodedAppInsightsKey)
-          
+
           #override ai endpoint if its available otherwise use default.
           if appInsightsEndpoint && !appInsightsEndpoint.nil? && !appInsightsEndpoint.empty?
             $log.info("AppInsightsUtility: Telemetry client uses overrided endpoint url : #{appInsightsEndpoint}")
@@ -110,11 +111,17 @@ def initializeUtility()
       end
     end
 
-    def getDockerInfo()
-      dockerInfo = DockerApiClient.dockerInfo
-      if (!dockerInfo.nil? && !dockerInfo.empty?)
-        @@CustomProperties["DockerVersion"] = dockerInfo["Version"]
-        #@@CustomProperties["DockerApiVersion"] = dockerInfo["ApiVersion"]
+    def getContainerRuntimeInfo()
+      containerRuntime = ENV[@@EnvContainerRuntime]
+      if !containerRuntime.nil? && !containerRuntime.empty?
+        # DockerVersion field holds either containerRuntime for non-docker or Dockerversion if its docker
+        @@CustomProperties["DockerVersion"] = containerRuntime
+        if containerRuntime.casecmp("docker") == 0
+          dockerInfo = DockerApiClient.dockerInfo
+          if (!dockerInfo.nil? && !dockerInfo.empty?)
+            @@CustomProperties["DockerVersion"] = dockerInfo["Version"]          
+          end
+        end
       end
     end
 
@@ -169,7 +176,7 @@ def sendExceptionTelemetry(errorStr, properties = nil)
         if @@CustomProperties.empty? || @@CustomProperties.nil?
           initializeUtility()
         elsif @@CustomProperties["DockerVersion"].nil?
-          getDockerInfo()
+          getContainerRuntimeInfo()
         end
         telemetryProps = {}
         # add common dimensions
@@ -193,7 +200,7 @@ def sendTelemetry(pluginName, properties)
         if @@CustomProperties.empty? || @@CustomProperties.nil?
           initializeUtility()
         elsif @@CustomProperties["DockerVersion"].nil?
-          getDockerInfo()
+          getContainerRuntimeInfo()
         end
         @@CustomProperties["Computer"] = properties["Computer"]
         sendHeartBeatEvent(pluginName)
@@ -213,7 +220,7 @@ def sendMetricTelemetry(metricName, metricValue, properties)
         if @@CustomProperties.empty? || @@CustomProperties.nil?
           initializeUtility()
         elsif @@CustomProperties["DockerVersion"].nil?
-          getDockerInfo()
+          getContainerRuntimeInfo()
         end
         telemetryProps = {}
         # add common dimensions

source/code/plugin/CAdvisorMetricsAPIClient.rb

@@ -73,6 +73,11 @@ def getNodeCapacityFromCAdvisor(winNode: nil)
       return getResponse(winNode, relativeUri)
     end
 
+    def getPodsFromCAdvisor(winNode: nil)
+      relativeUri = "/pods"
+      return getResponse(winNode, relativeUri)
+    end
+
     def getBaseCAdvisorUri(winNode)
         cAdvisorSecurePort = isCAdvisorOnSecurePort()