set mdsd limits (#1027)

* set mdsd limit using container memory limit

---------

Co-authored-by: Amol Agrawal <amagraw@microsoft.com>

Author: pfrcks

build/common/installer/scripts/tomlparser-agent-config.rb

@@ -83,6 +83,7 @@
 # Checking to see if container is not prometheus sidecar.
 # CONTAINER_TYPE is populated only for prometheus sidecar container.
 @containerType = ENV["CONTAINER_TYPE"]
+@containerMemoryLimitInBytes = ENV["CONTAINER_MEMORY_LIMIT_IN_BYTES"]
 
 @promFbitChunkSize = 0
 @promFbitBufferSize = 0
@@ -268,11 +269,12 @@ def populateSettingValuesFromConfigMap(parsedConfig)
           puts "Using config map value: require_ack_response = #{@requireAckResponse}"
         end
       end
-      # ama-logs daemonset only settings
-      if !@controllerType.nil? && !@controllerType.empty? && @controllerType.strip.casecmp(@daemonset) == 0 && @containerType.nil?
-        # mdsd settings
-        mdsd_config = parsedConfig[:agent_settings][:mdsd_config]
-        if !mdsd_config.nil?
+
+      # mdsd settings
+      mdsd_config = parsedConfig[:agent_settings][:mdsd_config]
+      if !mdsd_config.nil?
+        # ama-logs daemonset only settings
+        if !@controllerType.nil? && !@controllerType.empty? && @controllerType.strip.casecmp(@daemonset) == 0 && @containerType.nil?
           mdsdMonitoringMaxEventRate = mdsd_config[:monitoring_max_event_rate]
           if is_valid_number?(mdsdMonitoringMaxEventRate)
             @mdsdMonitoringMaxEventRate = mdsdMonitoringMaxEventRate.to_i
@@ -288,13 +290,6 @@ def populateSettingValuesFromConfigMap(parsedConfig)
             @mdsdUploadFrequencyInSeconds = mdsdUploadFrequencyInSeconds.to_i
             puts "Using config map value: upload_frequency_seconds  = #{@mdsdUploadFrequencyInSeconds}"
           end
-          mdsdBackPressureThresholdInMB = mdsd_config[:backpressure_memory_threshold_in_mb]
-          if is_valid_number?(mdsdBackPressureThresholdInMB) && mdsdBackPressureThresholdInMB.to_i > 100
-            @mdsdBackPressureThresholdInMB = mdsdBackPressureThresholdInMB.to_i
-            puts "Using config map value: backpressure_memory_threshold_in_mb  = #{@mdsdBackPressureThresholdInMB}"
-          else
-            puts "Ignoring mdsd backpressure limit. Check input values for correctness."
-          end
           mdsdCompressionLevel = mdsd_config[:compression_level]
           if is_number?(mdsdCompressionLevel) && mdsdCompressionLevel.to_i >= 0 && mdsdCompressionLevel.to_i < 10 # supported levels from 0 to 9
             @mdsdCompressionLevel = mdsdCompressionLevel.to_i
@@ -303,6 +298,14 @@ def populateSettingValuesFromConfigMap(parsedConfig)
             puts "Ignoring mdsd compression_level level since its not supported level. Check input values for correctness."
           end
         end
+
+        mdsdBackPressureThresholdInMB = mdsd_config[:backpressure_memory_threshold_in_mb]
+        if is_valid_number?(mdsdBackPressureThresholdInMB) && is_valid_number?(@containerMemoryLimitInBytes) && mdsdBackPressureThresholdInMB.to_i < (@containerMemoryLimitInBytes.to_i / 1048576) && mdsdBackPressureThresholdInMB.to_i > 100
+          @mdsdBackPressureThresholdInMB = mdsdBackPressureThresholdInMB.to_i
+          puts "Using config map value: backpressure_memory_threshold_in_mb  = #{@mdsdBackPressureThresholdInMB}"
+        else
+          puts "Ignoring mdsd backpressure limit. Check input values for correctness. Configmap value in mb: #{mdsdBackPressureThresholdInMB}, container limit in bytes: #{@containerMemoryLimitInBytes}"
+        end
       end
 
       prom_fbit_config = nil
@@ -443,7 +446,7 @@ def populateSettingValuesFromConfigMap(parsedConfig)
   end
 
   if @mdsdBackPressureThresholdInMB > 0
-    file.write("export MDSD_BACKPRESSURE_MONITOR_MEMORY_THRESHOLD_IN_MB=#{@mdsdBackPressureThresholdInMB}\n")
+    file.write("export BACKPRESSURE_THRESHOLD_IN_MB=#{@mdsdBackPressureThresholdInMB}\n")
   end
 
   if @mdsdCompressionLevel >= 0

build/common/installer/scripts/tomlparser-prom-agent-config.rb

@@ -16,11 +16,18 @@
 @waittime_port_25226 = 45
 @waittime_port_25228 = 120
 @waittime_port_25229 = 45
+@containerMemoryLimitInBytes = ENV["CONTAINER_MEMORY_LIMIT_IN_BYTES"]
+@mdsdBackPressureThresholdInMB = 0
 
 def is_number?(value)
   true if Integer(value) rescue false
 end
 
+# check if it is number and greater than 0
+def is_valid_number?(value)
+  return !value.nil? && is_number?(value) && value.to_i > 0
+end
+
 # check if it is a valid waittime
 def is_valid_waittime?(value, default)
   return !value.nil? && is_number?(value) && value.to_i >= default/2 && value.to_i <= 3*default
@@ -94,6 +101,18 @@ def populateSettingValuesFromConfigMap(parsedConfig)
         end
       end
 
+      # mdsd settings
+      mdsd_config = parsedConfig[:agent_settings][:mdsd_config]
+      if !mdsd_config.nil?
+        mdsdBackPressureThresholdInMB = mdsd_config[:backpressure_memory_threshold_in_mb]
+        if is_valid_number?(mdsdBackPressureThresholdInMB) && is_valid_number?(@containerMemoryLimitInBytes) && mdsdBackPressureThresholdInMB.to_i < (@containerMemoryLimitInBytes.to_i / 1048576) && mdsdBackPressureThresholdInMB.to_i > 100
+          @mdsdBackPressureThresholdInMB = mdsdBackPressureThresholdInMB.to_i
+          puts "Using config map value: backpressure_memory_threshold_in_mb  = #{@mdsdBackPressureThresholdInMB}"
+        else
+          puts "Ignoring mdsd backpressure limit. Check input values for correctness. Configmap value in mb: #{mdsdBackPressureThresholdInMB}, container limit in bytes: #{@containerMemoryLimitInBytes}"
+        end
+      end
+
     end
   rescue => errorStr
     puts "config::error:Exception while reading config settings for sidecar agent configuration setting - #{errorStr}, using defaults"
@@ -124,6 +143,11 @@ def populateSettingValuesFromConfigMap(parsedConfig)
   file.write("export WAITTIME_PORT_25226=#{@waittime_port_25226}\n")
   file.write("export WAITTIME_PORT_25228=#{@waittime_port_25228}\n")
   file.write("export WAITTIME_PORT_25229=#{@waittime_port_25229}\n")
+  
+  if @mdsdBackPressureThresholdInMB > 0
+    file.write("export BACKPRESSURE_THRESHOLD_IN_MB=#{@mdsdBackPressureThresholdInMB}\n")
+  end
+
   # Close file after writing all environment variables
   file.close
 else

charts/azuremonitor-containers/templates/ama-logs-daemonset-windows.yaml

@@ -85,6 +85,11 @@ spec:
          valueFrom:
             fieldRef:
               fieldPath: spec.nodeName
+       - name: CONTAINER_MEMORY_LIMIT_IN_BYTES
+         valueFrom:
+           resourceFieldRef:
+             containerName: ama-logs-windows
+             resource: limits.memory
        - name: NODE_IP
          valueFrom:
             fieldRef:

charts/azuremonitor-containers/templates/ama-logs-daemonset.yaml

@@ -83,6 +83,11 @@ spec:
          valueFrom:
             fieldRef:
               fieldPath: status.hostIP
+       - name: CONTAINER_MEMORY_LIMIT_IN_BYTES
+         valueFrom:
+           resourceFieldRef:
+             containerName: ama-logs
+             resource: limits.memory
        {{- if not (empty .Values.Azure.Extension.Name) }}
        - name: ARC_K8S_EXTENSION_NAME
          value: {{ .Values.Azure.Extension.Name | quote }}
@@ -210,6 +215,11 @@ spec:
          valueFrom:
             fieldRef:
               fieldPath: status.hostIP
+       - name: CONTAINER_MEMORY_LIMIT_IN_BYTES
+         valueFrom:
+           resourceFieldRef:
+             containerName: ama-logs-prometheus
+             resource: limits.memory
        - name: ISTEST
          value: {{ .Values.amalogs.ISTEST | quote }}
        - name: HOSTNAME

charts/azuremonitor-containers/templates/ama-logs-deployment.yaml

@@ -58,6 +58,11 @@ spec:
            resourceFieldRef:
              containerName: ama-logs
              resource: limits.cpu
+       - name: CONTAINER_MEMORY_LIMIT_IN_BYTES
+         valueFrom:
+           resourceFieldRef:
+             containerName: ama-logs
+             resource: limits.memory
        {{- if ne .Values.amalogs.env.clusterId "<your_cluster_id>" }}
        - name: AKS_RESOURCE_ID
          value: {{ .Values.amalogs.env.clusterId | quote }}

kubernetes/linux/main.sh

@@ -855,6 +855,28 @@ else
 fi
 source ~/.bashrc
 
+# manually set backpressure value using container limit only when neither backpressure or fbit tail buffer is provided through configmap
+if [ -n "${BACKPRESSURE_THRESHOLD_IN_MB}" ]; then
+      export MDSD_BACKPRESSURE_MONITOR_MEMORY_THRESHOLD_IN_MB=${BACKPRESSURE_THRESHOLD_IN_MB}
+      echo "export MDSD_BACKPRESSURE_MONITOR_MEMORY_THRESHOLD_IN_MB=$MDSD_BACKPRESSURE_MONITOR_MEMORY_THRESHOLD_IN_MB" >> ~/.bashrc
+      echo "Setting MDSD backpressure threshold from configmap: ${MDSD_BACKPRESSURE_MONITOR_MEMORY_THRESHOLD_IN_MB} MB"
+      source ~/.bashrc
+elif [ -z "${FBIT_TAIL_MEM_BUF_LIMIT}" ]; then
+      if [ -n "${CONTAINER_MEMORY_LIMIT_IN_BYTES}" ]; then
+            echo "Container limit in bytes: ${CONTAINER_MEMORY_LIMIT_IN_BYTES}"
+            limit_in_mebibytes=$((CONTAINER_MEMORY_LIMIT_IN_BYTES / 1048576))
+
+            export MDSD_BACKPRESSURE_MONITOR_MEMORY_THRESHOLD_IN_MB=$((limit_in_mebibytes * 50 / 100))
+            echo "export MDSD_BACKPRESSURE_MONITOR_MEMORY_THRESHOLD_IN_MB=$MDSD_BACKPRESSURE_MONITOR_MEMORY_THRESHOLD_IN_MB" >> ~/.bashrc
+            echo "Setting MDSD backpressure threshold as 50 percent of container limit: ${MDSD_BACKPRESSURE_MONITOR_MEMORY_THRESHOLD_IN_MB} MB"
+            source ~/.bashrc
+      else
+            echo "Container limit not found. Not setting mdsd backpressure threshold"
+      fi
+else
+      echo "MDSD backpressure threshold not set since tail_mem_buf_limit_megabytes is used in configmap. Use backpressure_memory_threshold_in_mb in configmap to set it."
+fi
+
 if [ "${CONTAINER_TYPE}" == "PrometheusSidecar" ]; then
     if [ "${MUTE_PROM_SIDECAR}" != "true" ]; then
       echo "starting mdsd with mdsd-port=26130, fluentport=26230 and influxport=26330 in sidecar container..."

source/plugins/go/src/telemetry.go

@@ -185,6 +185,11 @@ func SendContainerLogPluginMetrics(telemetryPushIntervalProperty string) {
 				telemetryDimensions["PromFbitBufferSize"] = os.Getenv("AZMON_SIDECAR_FBIT_BUFFER_SIZE")
 				telemetryDimensions["PromFbitMemBufLimit"] = os.Getenv("AZMON_SIDECAR_FBIT_MEM_BUF_LIMIT")
 
+				mdsdBackPressureThresholdInMB := os.Getenv("MDSD_BACKPRESSURE_MONITOR_MEMORY_THRESHOLD_IN_MB")
+				if mdsdBackPressureThresholdInMB != "" {
+					telemetryDimensions["mdsdBackPressureThresholdInMB"] = mdsdBackPressureThresholdInMB
+				}
+
 				SendEvent(eventNameCustomPrometheusSidecarHeartbeat, telemetryDimensions)
 
 			} else {

source/plugins/ruby/in_kube_nodes.rb

@@ -50,6 +50,7 @@ def initialize(is_unit_test_mode = nil, kubernetesApiClient = nil,
       @@rsPromMonitorPodsFieldSelectorLength = @env["TELEMETRY_RS_PROM_FIELD_SELECTOR_LENGTH"]
       @@collectAllKubeEvents = @env["AZMON_CLUSTER_COLLECT_ALL_KUBE_EVENTS"]
       @@osmNamespaceCount = @env["TELEMETRY_OSM_CONFIGURATION_NAMESPACES_COUNT"]
+      @@mdsdBackPressureThresholdInMB = @env["MDSD_BACKPRESSURE_MONITOR_MEMORY_THRESHOLD_IN_MB"]
 
       @ContainerNodeInventoryTag = "oneagent.containerInsights.CONTAINER_NODE_INVENTORY_BLOB"
       @insightsMetricsTag = "oneagent.containerInsights.INSIGHTS_METRICS_BLOB"
@@ -393,6 +394,12 @@ def parse_and_emit_records(nodeInventory, batchTime = Time.utc.iso8601)
               if (File.file?(@@osmConfigMountPath))
                 properties["osmNamespaceCount"] = @@osmNamespaceCount
               end
+
+              # telemetry about mdsd backpressure limits for replicaset
+              if (!@@mdsdBackPressureThresholdInMB.nil?) && (!@@mdsdBackPressureThresholdInMB.empty?)
+                properties["mdsdBackPressureThresholdInMB"] = @@mdsdBackPressureThresholdInMB
+              end
+
               @applicationInsightsUtility.sendMetricTelemetry("NodeCoreCapacity", capacityInfo["cpu"], properties)
               telemetrySent = true
             rescue => errorStr