DP-1537: Limit Influx db volume getting huge overtime

Dockerfile

@@ -1,15 +1,21 @@
-ARG TELEGRAF_VERSION="1.24-alpine"
+ARG TELEGRAF_VERSION="1.30-alpine"
 FROM telegraf:${TELEGRAF_VERSION}
 
+ARG ENVIRONMENT="production"
+
 # Labels
 LABEL description="MOV.AI metrics collector"
 LABEL maintainer="[email protected]"
 LABEL movai="telegraf"
-LABEL environment="develop"
+LABEL environment=$ENVIRONMENT
 
-COPY files/telegraf.conf /etc/telegraf/telegraf.conf
+# Copy configuration files
+COPY files/telegraf_debug.conf /etc/telegraf/telegraf_debug.conf
+COPY files/telegraf_production.conf /etc/telegraf/telegraf_production.conf
 COPY files/entrypoint.sh /entrypoint.sh
 
-RUN chmod o+w /etc/telegraf/telegraf.conf
+ENV TELEGRAF_CONFIG_LEVEL $ENVIRONMENT
 
-# Rest is upstream
+RUN chmod o+w /etc/telegraf/telegraf_debug.conf /etc/telegraf/telegraf_production.conf \
+    && rm -f /etc/telegraf/telegraf.conf \
+    && chmod +x /entrypoint.sh

files/entrypoint.sh

@@ -1,5 +1,6 @@
 #!/bin/sh
 set -e
+TELEGRAF_CONFIG_PATH=/etc/telegraf/telegraf_${TELEGRAF_CONFIG_LEVEL}.conf
 
 # Check if linux_cpu inputs required files exist
 enable_plugin_cpufreq=false
@@ -36,16 +37,18 @@ if [ "$enable_plugin_thermalthrottle" = "true" ]; then
     metrics="${metrics},\"thermal\""
 fi
 
-if [ -n "$metrics" ] && ! grep linux_cpu /etc/telegraf/telegraf.conf -q; then
+if [ -n "$metrics" ] && ! grep linux_cpu $TELEGRAF_CONFIG_PATH -q; then
     # Include the configuration for the [[inputs.linux_cpu]] plugin
-    cat <<EOF >> /etc/telegraf/telegraf.conf
+    cat <<EOF >> $TELEGRAF_CONFIG_PATH
 
 [[inputs.linux_cpu]]
   host_sys = "/hostfs/sys"
   metrics = [$metrics]
 EOF
 fi
 
+ln -sf $TELEGRAF_CONFIG_PATH /etc/telegraf/telegraf.conf
+
 if [ "${1:0:1}" = '-' ]; then
     set -- telegraf "$@"
 fi

files/telegraf_debug.conf

@@ -0,0 +1,59 @@
+[global_tags]
+[agent]
+  interval = "10s"
+  round_interval = true
+  metric_batch_size = 1000
+  metric_buffer_limit = 10000
+  collection_jitter = "1s"
+  flush_interval = "10s"
+  flush_jitter = "0s"
+  precision = "0s"
+  debug = false
+  quiet = false
+  hostname = "{{.Hostname}}"
+  omit_hostname = false
+[[outputs.influxdb]]
+  urls = ["udp://influxdb:9096"] # required
+  database = "telegraf" # required
+  retention_policy = ""
+  write_consistency = "any"
+  timeout = "5s"
+  udp_payload = "61440B"
+[[inputs.cpu]]
+  percpu = true
+  totalcpu = true
+  collect_cpu_time = false
+  report_active = false
+  core_tags = false
+[[inputs.disk]]
+  ignore_fs = ["tmpfs", "devtmpfs", "devfs", "iso9660", "overlay", "aufs", "squashfs", "nsfs", "fuse.portal", "fuse.gvfsd-fuse", "fuse.lxcfs", "fuse.dockerfs", "fuse", "hugetlbfs", "mqueue", "vfat"]
+[[inputs.diskio]]
+[[inputs.kernel]]
+[[inputs.mem]]
+[[inputs.processes]]
+[[inputs.swap]]
+[[inputs.system]]
+[[inputs.docker]]
+  endpoint = "unix:///var/run/docker.sock"
+  timeout = "5s"
+  perdevice = false
+  perdevice_include = ["cpu", "blkio", "network"]
+[[inputs.influxdb]]
+  urls = [
+    "http://influxdb:8086/debug/vars"
+  ]
+  timeout = "5s"
+[[inputs.internal]]
+  collect_memstats = true
+[[inputs.linux_sysctl_fs]]
+[[inputs.netstat]]
+[[inputs.procstat]]
+  pattern = "."
+  pid_finder = "native"
+  tag_with = ["cmdline"]
+[[inputs.wireless]]
+  host_proc = "/hostfs/proc/1"
+[[inputs.docker_log]]
+    endpoint = "unix:///var/run/docker.sock"
+[[inputs.redis]]
+  servers = ["tcp://redis-local:6379","tcp://redis-master:6379"]