| 1 | #!/usr/bin/env bash |
| 2 | # SPDX-License-Identifier: GPL-3.0-or-later |
| 3 | # |
| 4 | # Script to find a better name for cgroups |
| 5 | # |
| 6 | #shellcheck disable=SC2001 |
| 7 | |
| 8 | export PATH="${PATH}:/sbin:/usr/sbin:/usr/local/sbin:@sbindir_POST@" |
| 9 | export LC_ALL=C |
| 10 | |
| 11 | cmd_line="'${0}' $(printf "'%s' " "${@}")" |
| 12 | |
| 13 | # ----------------------------------------------------------------------------- |
| 14 | # logging |
| 15 | |
| 16 | PROGRAM_NAME="$(basename "${0}")" |
| 17 | |
| 18 | # these should be the same with syslog() priorities |
| 19 | NDLP_EMERG=0 # system is unusable |
| 20 | NDLP_ALERT=1 # action must be taken immediately |
| 21 | NDLP_CRIT=2 # critical conditions |
| 22 | NDLP_ERR=3 # error conditions |
| 23 | NDLP_WARN=4 # warning conditions |
| 24 | NDLP_NOTICE=5 # normal but significant condition |
| 25 | NDLP_INFO=6 # informational |
| 26 | NDLP_DEBUG=7 # debug-level messages |
| 27 | |
| 28 | # the max (numerically) log level we will log |
| 29 | LOG_LEVEL=$NDLP_INFO |
| 30 | |
| 31 | set_log_min_priority() { |
| 32 | case "${NETDATA_LOG_LEVEL,,}" in |
| 33 | "emerg" | "emergency") |
| 34 | LOG_LEVEL=$NDLP_EMERG |
| 35 | ;; |
| 36 | |
| 37 | "alert") |
| 38 | LOG_LEVEL=$NDLP_ALERT |
| 39 | ;; |
| 40 | |
| 41 | "crit" | "critical") |
| 42 | LOG_LEVEL=$NDLP_CRIT |
| 43 | ;; |
| 44 | |
| 45 | "err" | "error") |
| 46 | LOG_LEVEL=$NDLP_ERR |
| 47 | ;; |
| 48 | |
| 49 | "warn" | "warning") |
| 50 | LOG_LEVEL=$NDLP_WARN |
| 51 | ;; |
| 52 | |
| 53 | "notice") |
| 54 | LOG_LEVEL=$NDLP_NOTICE |
| 55 | ;; |
| 56 | |
| 57 | "info") |
| 58 | LOG_LEVEL=$NDLP_INFO |
| 59 | ;; |
| 60 | |
| 61 | "debug") |
| 62 | LOG_LEVEL=$NDLP_DEBUG |
| 63 | ;; |
| 64 | esac |
| 65 | } |
| 66 | |
| 67 | set_log_min_priority |
| 68 | |
| 69 | log() { |
| 70 | local level="${1}" |
| 71 | shift 1 |
| 72 | |
| 73 | [[ -n "$level" && -n "$LOG_LEVEL" && "$level" -gt "$LOG_LEVEL" ]] && return |
| 74 | |
| 75 | systemd-cat-native --log-as-netdata << EOFLOG |
| 76 | INVOCATION_ID=${NETDATA_INVOCATION_ID} |
| 77 | SYSLOG_IDENTIFIER=${PROGRAM_NAME} |
| 78 | PRIORITY=${level} |
| 79 | THREAD_TAG=cgroup-name |
| 80 | ND_LOG_SOURCE=collector |
| 81 | ND_REQUEST=${cmd_line} |
| 82 | MESSAGE=${*//$'\n'/\\n} |
| 83 | |
| 84 | EOFLOG |
| 85 | # AN EMPTY LINE IS NEEDED ABOVE |
| 86 | } |
| 87 | |
| 88 | info() { |
| 89 | log "$NDLP_INFO" "${@}" |
| 90 | } |
| 91 | |
| 92 | warning() { |
| 93 | log "$NDLP_WARN" "${@}" |
| 94 | } |
| 95 | |
| 96 | error() { |
| 97 | log "$NDLP_ERR" "${@}" |
| 98 | } |
| 99 | |
| 100 | fatal() { |
| 101 | log "$NDLP_ALERT" "${@}" |
| 102 | exit 1 |
| 103 | } |
| 104 | |
| 105 | debug() { |
| 106 | log "$NDLP_DEBUG" "${@}" |
| 107 | } |
| 108 | |
| 109 | # ----------------------------------------------------------------------------- |
| 110 | |
| 111 | function parse_docker_like_inspect_output() { |
| 112 | local output="${1}" |
| 113 | eval "$(grep -E "^(NOMAD_NAMESPACE|NOMAD_JOB_NAME|NOMAD_TASK_NAME|NOMAD_SHORT_ALLOC_ID|CONT_NAME|IMAGE_NAME)=" <<< "$output")" |
| 114 | if [ -n "$NOMAD_NAMESPACE" ] && [ -n "$NOMAD_JOB_NAME" ] && [ -n "$NOMAD_TASK_NAME" ] && [ -n "$NOMAD_SHORT_ALLOC_ID" ]; then |
| 115 | NAME="${NOMAD_NAMESPACE}-${NOMAD_JOB_NAME}-${NOMAD_TASK_NAME}-${NOMAD_SHORT_ALLOC_ID}" |
| 116 | else |
| 117 | NAME=$(echo "${CONT_NAME}" | sed 's|^/||') |
| 118 | fi |
| 119 | if [ -n "${IMAGE_NAME}" ]; then |
| 120 | LABELS="image=\"${IMAGE_NAME}\"" |
| 121 | fi |
| 122 | |
| 123 | while IFS= read -r line; do |
| 124 | if [[ $line == LABEL_netdata.cloud/* ]]; then |
| 125 | IFS="=" read -r lname lval <<< "$line" |
| 126 | lname=${lname#LABEL_} |
| 127 | if [ -n "$LABELS" ]; then |
| 128 | LABELS="$LABELS,${lname}=\"${lval}\"" |
| 129 | else |
| 130 | LABELS="${lname}=\"${lval}\"" |
| 131 | fi |
| 132 | fi |
| 133 | done <<< "$output" |
| 134 | } |
| 135 | |
| 136 | function docker_like_get_name_command() { |
| 137 | local command="${1}" |
| 138 | local id="${2}" |
| 139 | # shellcheck disable=SC2016 |
| 140 | if OUTPUT="$(${command} inspect --format='{{range .Config.Env}}{{println .}}{{end}}{{range $key, $value := .Config.Labels}}LABEL_{{$key}}={{$value}}{{println}}{{end}}IMAGE_NAME={{.Config.Image}}{{println}}CONT_NAME={{.Name}}' "${id}")" && |
| 141 | [ -n "$OUTPUT" ]; then |
| 142 | parse_docker_like_inspect_output "$OUTPUT" |
| 143 | fi |
| 144 | return 0 |
| 145 | } |
| 146 | |
| 147 | function docker_like_get_name_api() { |
| 148 | local host_var="${1}" |
| 149 | local host="${!host_var}" |
| 150 | local container_id="${2}" |
| 151 | local path="/containers/${container_id}/json" |
| 152 | |
| 153 | if [ -z "${host}" ]; then |
| 154 | warning "No ${host_var} is set" |
| 155 | return 1 |
| 156 | fi |
| 157 | |
| 158 | if ! command -v jq > /dev/null 2>&1; then |
| 159 | warning "Can't find jq command line tool. jq is required for netdata to retrieve container name using ${host} API, falling back to docker ps" |
| 160 | return 1 |
| 161 | fi |
| 162 | |
| 163 | if [[ "${host}" =~ ^([a-z]+)://(.*) ]]; then |
| 164 | address="${BASH_REMATCH[2]}" |
| 165 | else |
| 166 | address="${host}" |
| 167 | fi |
| 168 | |
| 169 | if [ -S "${address}" ]; then |
| 170 | info "Running API command: curl --unix-socket \"${address}\" http://localhost${path}" |
| 171 | JSON=$(curl -sS --unix-socket "${address}" "http://localhost${path}") |
| 172 | else |
| 173 | info "Running API command: curl \"${address}${path}\"" |
| 174 | JSON=$(curl -sS "${address}${path}") |
| 175 | fi |
| 176 | |
| 177 | if OUTPUT=$(echo "${JSON}" | jq -r '.Config.Env[]?, "CONT_NAME=\(.Name)", "IMAGE_NAME=\(.Config.Image)", (.Config.Labels | to_entries[] | "LABEL_\(.key)=\(.value)")') && [ -n "$OUTPUT" ]; then |
| 178 | parse_docker_like_inspect_output "$OUTPUT" |
| 179 | fi |
| 180 | |
| 181 | return 0 |
| 182 | } |
| 183 | |
| 184 | # get_lbl_val returns the value for the label with the given name. |
| 185 | # Returns "null" string if the label doesn't exist. |
| 186 | # Expected labels format: 'name="value",...'. |
| 187 | function get_lbl_val() { |
| 188 | local labels want_name |
| 189 | labels="${1}" |
| 190 | want_name="${2}" |
| 191 | |
| 192 | IFS=, read -ra labels <<< "$labels" |
| 193 | |
| 194 | local lname lval |
| 195 | for l in "${labels[@]}"; do |
| 196 | IFS="=" read -r lname lval <<< "$l" |
| 197 | if [ "$want_name" = "$lname" ] && [ -n "$lval" ]; then |
| 198 | echo "${lval:1:-1}" # trim " |
| 199 | return 0 |
| 200 | fi |
| 201 | done |
| 202 | |
| 203 | echo "null" |
| 204 | return 1 |
| 205 | } |
| 206 | |
| 207 | function add_lbl_prefix() { |
| 208 | local orig_labels prefix |
| 209 | orig_labels="${1}" |
| 210 | prefix="${2}" |
| 211 | |
| 212 | IFS=, read -ra labels <<< "$orig_labels" |
| 213 | |
| 214 | local new_labels |
| 215 | for l in "${labels[@]}"; do |
| 216 | new_labels+="${prefix}${l}," |
| 217 | done |
| 218 | |
| 219 | echo "${new_labels:0:-1}" # trim last ',' |
| 220 | } |
| 221 | |
| 222 | function remove_lbl() { |
| 223 | local orig_labels lbl_name |
| 224 | orig_labels="${1}" |
| 225 | lbl_name="${2}" |
| 226 | |
| 227 | IFS=, read -ra labels <<< "$orig_labels" |
| 228 | |
| 229 | local new_labels |
| 230 | for l in "${labels[@]}"; do |
| 231 | IFS="=" read -r lname lval <<< "$l" |
| 232 | [ "$lbl_name" != "$lname" ] && new_labels+="${l}," |
| 233 | done |
| 234 | |
| 235 | echo "${new_labels:0:-1}" # trim last ',' |
| 236 | } |
| 237 | |
| 238 | function k8s_is_pause_container() { |
| 239 | local cgroup_path="${1}" |
| 240 | |
| 241 | local file |
| 242 | if [ -d "${NETDATA_HOST_PREFIX}/sys/fs/cgroup/cpuacct" ]; then |
| 243 | file="${NETDATA_HOST_PREFIX}/sys/fs/cgroup/cpuacct/$cgroup_path/cgroup.procs" |
| 244 | else |
| 245 | file="${NETDATA_HOST_PREFIX}/sys/fs/cgroup/$cgroup_path/cgroup.procs" |
| 246 | fi |
| 247 | |
| 248 | [ ! -f "$file" ] && return 1 |
| 249 | |
| 250 | local procs |
| 251 | IFS= read -rd' ' procs 2> /dev/null < "$file" |
| 252 | #shellcheck disable=SC2206 |
| 253 | procs=($procs) |
| 254 | |
| 255 | [ "${#procs[@]}" -ne 1 ] && return 1 |
| 256 | |
| 257 | IFS= read -r comm 2> /dev/null < "/proc/${procs[0]}/comm" |
| 258 | |
| 259 | [ "$comm" == "pause" ] |
| 260 | return |
| 261 | } |
| 262 | |
| 263 | function k8s_gcp_get_cluster_name() { |
| 264 | local header url id loc name |
| 265 | header="Metadata-Flavor: Google" |
| 266 | url="http://metadata/computeMetadata/v1" |
| 267 | if id=$(curl --fail -s -m 3 --noproxy "*" -H "$header" "$url/project/project-id") && |
| 268 | loc=$(curl --fail -s -m 3 --noproxy "*" -H "$header" "$url/instance/attributes/cluster-location") && |
| 269 | name=$(curl --fail -s -m 3 --noproxy "*" -H "$header" "$url/instance/attributes/cluster-name") && |
| 270 | [ -n "$id" ] && [ -n "$loc" ] && [ -n "$name" ]; then |
| 271 | echo "gke_${id}_${loc}_${name}" |
| 272 | return 0 |
| 273 | fi |
| 274 | return 1 |
| 275 | } |
| 276 | |
| 277 | # k8s_get_kubepod_name resolves */kubepods/* cgroup name. |
| 278 | # pod level cgroup name format: 'pod_<namespace>_<pod_name>' |
| 279 | # container level cgroup name format: 'cntr_<namespace>_<pod_name>_<container_name>' |
| 280 | function k8s_get_kubepod_name() { |
| 281 | # GKE /sys/fs/cgroup/*/ (cri=docker, cgroups=v1): |
| 282 | # |-- kubepods |
| 283 | # | |-- burstable |
| 284 | # | | |-- pod98cee708-023b-11eb-933d-42010a800193 |
| 285 | # | | | |-- 922161c98e6ea450bf665226cdc64ca2aa3e889934c2cff0aec4325f8f78ac03 |
| 286 | # | `-- pode314bbac-d577-11ea-a171-42010a80013b |
| 287 | # | |-- 7d505356b04507de7b710016d540b2759483ed5f9136bb01a80872b08f771930 |
| 288 | # |
| 289 | # GKE /sys/fs/cgroup/*/ (cri=containerd, cgroups=v1): |
| 290 | # |-- kubepods.slice |
| 291 | # | |-- kubepods-besteffort.slice |
| 292 | # | | |-- kubepods-besteffort-pode1465238_4518_4c21_832f_fd9f87033dad.slice |
| 293 | # | | | |-- cri-containerd-66be9b2efdf4d85288c319b8c1a2f50d2439b5617e36f45d9d0d0be1381113be.scope |
| 294 | # | `-- kubepods-pod91f5b561_369f_4103_8015_66391059996a.slice |
| 295 | # | |-- cri-containerd-24c53b774a586f06abc058619b47f71d9d869ac50c92898adbd199106fd0aaeb.scope |
| 296 | # |
| 297 | # GKE /sys/fs/cgroup/*/ (cri=crio, cgroups=v1): |
| 298 | # |-- kubepods.slice |
| 299 | # | |-- kubepods-besteffort.slice |
| 300 | # | | |-- kubepods-besteffort-podad412dfe_3589_4056_965a_592356172968.slice |
| 301 | # | | | |-- crio-77b019312fd9825828b70214b2c94da69c30621af2a7ee06f8beace4bc9439e5.scope |
| 302 | # |
| 303 | # Minikube (v1.8.2) /sys/fs/cgroup/*/ (cri=docker, cgroups=v1): |
| 304 | # |-- kubepods.slice |
| 305 | # | |-- kubepods-besteffort.slice |
| 306 | # | | |-- kubepods-besteffort-pod10fb5647_c724_400c_b9cc_0e6eae3110e7.slice |
| 307 | # | | | |-- docker-36e5eb5056dfdf6dbb75c0c44a1ecf23217fe2c50d606209d8130fcbb19fb5a7.scope |
| 308 | # |
| 309 | # kind v0.14.0 |
| 310 | # |-- kubelet.slice |
| 311 | # | |-- kubelet-kubepods.slice |
| 312 | # | | |-- kubelet-kubepods-besteffort.slice |
| 313 | # | | | |-- kubelet-kubepods-besteffort-pod7881ed9e_c63e_4425_b5e0_ac55a08ae939.slice |
| 314 | # | | | | |-- cri-containerd-00c7939458bffc416bb03451526e9fde13301d6654cfeadf5b4964a7fb5be1a9.scope |
| 315 | # |
| 316 | # NOTE: cgroups plugin |
| 317 | # - uses '_' to join dir names (so it is <parent>_<child>_<child>_...) |
| 318 | # - replaces '.' with '-' |
| 319 | |
| 320 | local fn="${FUNCNAME[0]}" |
| 321 | local cgroup_path="${1}" |
| 322 | local id="${2}" |
| 323 | |
| 324 | if [[ ! $id =~ ^.*kubepods.* ]]; then |
| 325 | warning "${fn}: '${id}' is not kubepod cgroup." |
| 326 | return 1 |
| 327 | fi |
| 328 | |
| 329 | local clean_id="$id" |
| 330 | clean_id=${clean_id//.slice/} |
| 331 | clean_id=${clean_id//.scope/} |
| 332 | |
| 333 | local name pod_uid cntr_id |
| 334 | if [[ $clean_id == "kubepods" ]]; then |
| 335 | name="$clean_id" |
| 336 | elif [[ $clean_id =~ .+(besteffort|burstable|guaranteed)$ ]]; then |
| 337 | # kubepods_<QOS_CLASS> |
| 338 | # kubepods_kubepods-<QOS_CLASS> |
| 339 | name=${clean_id//-/_} |
| 340 | name=${name/#kubepods_kubepods/kubepods} |
| 341 | elif [[ $clean_id =~ .+pod[a-f0-9_-]+_(docker|crio|cri-containerd)-([a-f0-9]+)$ ]]; then |
| 342 | # ...pod<POD_UID>_(docker|crio|cri-containerd)-<CONTAINER_ID> (POD_UID w/ "_") |
| 343 | cntr_id=${BASH_REMATCH[2]} |
| 344 | elif [[ $clean_id =~ .+pod[a-f0-9-]+_([a-f0-9]+)$ ]]; then |
| 345 | # ...pod<POD_UID>_<CONTAINER_ID> |
| 346 | cntr_id=${BASH_REMATCH[1]} |
| 347 | elif [[ $clean_id =~ .+pod([a-f0-9_-]+)$ ]]; then |
| 348 | # ...pod<POD_UID> (POD_UID w/ and w/o "_") |
| 349 | pod_uid=${BASH_REMATCH[1]} |
| 350 | pod_uid=${pod_uid//_/-} |
| 351 | fi |
| 352 | |
| 353 | if [ -n "$name" ]; then |
| 354 | echo "$name" |
| 355 | return 0 |
| 356 | fi |
| 357 | |
| 358 | if [ -z "$pod_uid" ] && [ -z "$cntr_id" ]; then |
| 359 | warning "${fn}: can't extract pod_uid or container_id from the cgroup '$id'." |
| 360 | return 3 |
| 361 | fi |
| 362 | |
| 363 | [ -n "$pod_uid" ] && info "${fn}: cgroup '$id' is a pod(uid:$pod_uid)" |
| 364 | [ -n "$cntr_id" ] && info "${fn}: cgroup '$id' is a container(id:$cntr_id)" |
| 365 | |
| 366 | if [ -n "$cntr_id" ] && k8s_is_pause_container "$cgroup_path"; then |
| 367 | return 3 |
| 368 | fi |
| 369 | |
| 370 | if ! command -v jq > /dev/null 2>&1; then |
| 371 | warning "${fn}: 'jq' command not available." |
| 372 | return 1 |
| 373 | fi |
| 374 | |
| 375 | local tmp_kube_cluster_name="${TMPDIR:-"/tmp"}/netdata-cgroups-k8s-cluster-name" |
| 376 | local tmp_kube_system_ns_uid_file="${TMPDIR:-"/tmp"}/netdata-cgroups-kubesystem-uid" |
| 377 | local tmp_kube_containers_file="${TMPDIR:-"/tmp"}/netdata-cgroups-containers" |
| 378 | |
| 379 | local kube_cluster_name |
| 380 | local kube_system_uid |
| 381 | local labels |
| 382 | |
| 383 | if [ -n "$cntr_id" ] && |
| 384 | [ -f "$tmp_kube_cluster_name" ] && |
| 385 | [ -f "$tmp_kube_system_ns_uid_file" ] && |
| 386 | [ -f "$tmp_kube_containers_file" ] && |
| 387 | labels=$(grep "$cntr_id" "$tmp_kube_containers_file" 2> /dev/null); then |
| 388 | IFS= read -r kube_system_uid 2> /dev/null < "$tmp_kube_system_ns_uid_file" |
| 389 | IFS= read -r kube_cluster_name 2> /dev/null < "$tmp_kube_cluster_name" |
| 390 | else |
| 391 | IFS= read -r kube_system_uid 2> /dev/null < "$tmp_kube_system_ns_uid_file" |
| 392 | IFS= read -r kube_cluster_name 2> /dev/null < "$tmp_kube_cluster_name" |
| 393 | [ -z "$kube_cluster_name" ] && ! kube_cluster_name=$(k8s_gcp_get_cluster_name) && kube_cluster_name="unknown" |
| 394 | |
| 395 | local kube_system_ns |
| 396 | local pods |
| 397 | |
| 398 | if [ -n "${KUBERNETES_SERVICE_HOST}" ] && [ -n "${KUBERNETES_PORT_443_TCP_PORT}" ]; then |
| 399 | local token header host url |
| 400 | token="$(< /var/run/secrets/kubernetes.io/serviceaccount/token)" |
| 401 | header="Authorization: Bearer $token" |
| 402 | host="$KUBERNETES_SERVICE_HOST:$KUBERNETES_PORT_443_TCP_PORT" |
| 403 | |
| 404 | if [ -z "$kube_system_uid" ]; then |
| 405 | url="https://$host/api/v1/namespaces/kube-system" |
| 406 | # FIX: check HTTP response code |
| 407 | if ! kube_system_ns=$(curl --fail -sSk -H "$header" "$url" 2>&1); then |
| 408 | warning "${fn}: error on curl '${url}': ${kube_system_ns}." |
| 409 | fi |
| 410 | fi |
| 411 | |
| 412 | local url |
| 413 | if [ -n "${USE_KUBELET_FOR_PODS_METADATA}" ]; then |
| 414 | url="${KUBELET_URL:-https://localhost:10250}/pods" |
| 415 | else |
| 416 | url="https://$host/api/v1/pods" |
| 417 | [ -n "$MY_NODE_NAME" ] && url+="?fieldSelector=spec.nodeName==$MY_NODE_NAME" |
| 418 | fi |
| 419 | |
| 420 | # FIX: check HTTP response code |
| 421 | if ! pods=$(curl --fail -sSk -H "$header" "$url" 2>&1); then |
| 422 | warning "${fn}: error on curl '${url}': ${pods}." |
| 423 | return 1 |
| 424 | fi |
| 425 | elif ps -C kubelet > /dev/null 2>&1 && command -v kubectl > /dev/null 2>&1; then |
| 426 | if [ -z "$kube_system_uid" ]; then |
| 427 | if ! kube_system_ns=$(kubectl --kubeconfig="$KUBE_CONFIG" get namespaces kube-system -o json 2>&1); then |
| 428 | warning "${fn}: error on 'kubectl': ${kube_system_ns}." |
| 429 | fi |
| 430 | fi |
| 431 | |
| 432 | [[ -z ${KUBE_CONFIG+x} ]] && KUBE_CONFIG="/etc/kubernetes/admin.conf" |
| 433 | if ! pods=$(kubectl --kubeconfig="$KUBE_CONFIG" get pods --all-namespaces -o json 2>&1); then |
| 434 | warning "${fn}: error on 'kubectl': ${pods}." |
| 435 | return 1 |
| 436 | fi |
| 437 | else |
| 438 | warning "${fn}: not inside the k8s cluster and 'kubectl' command not available." |
| 439 | return 1 |
| 440 | fi |
| 441 | |
| 442 | if [ -n "$kube_system_ns" ] && ! kube_system_uid=$(jq -r '.metadata.uid' <<< "$kube_system_ns" 2>&1); then |
| 443 | warning "${fn}: error on 'jq' parse kube_system_ns: ${kube_system_uid}." |
| 444 | fi |
| 445 | |
| 446 | local jq_filter |
| 447 | jq_filter+='.items[] | "' |
| 448 | jq_filter+='namespace=\"\(.metadata.namespace)\",' |
| 449 | jq_filter+='pod_name=\"\(.metadata.name)\",' |
| 450 | jq_filter+='pod_uid=\"\(.metadata.uid)\",' |
| 451 | jq_filter+='\((.metadata.annotations // {}) | to_entries | map(select(.key | startswith("netdata.cloud/")) | .key+"=\""+.value+"\"") | join(",") | if length > 0 then .+"," else . end)' |
| 452 | jq_filter+='\((.metadata.ownerReferences[]? | select(.controller==true) | "controller_kind=\""+.kind+"\",controller_name=\""+.name+"\",") // "")' |
| 453 | jq_filter+='node_name=\"\(.spec.nodeName)\",' |
| 454 | jq_filter+='" + ' |
| 455 | jq_filter+='(.status.containerStatuses[]? | "' |
| 456 | jq_filter+='container_name=\"\(.name)\",' |
| 457 | jq_filter+='container_id=\"\(.containerID)\"' |
| 458 | jq_filter+='") | ' |
| 459 | jq_filter+='sub("(docker|cri-o|containerd)://";"")' # containerID: docker://a346da9bc0e3eaba6b295f64ac16e02f2190db2cef570835706a9e7a36e2c722 |
| 460 | |
| 461 | local containers |
| 462 | if ! containers=$(jq -r "${jq_filter}" <<< "$pods" 2>&1); then |
| 463 | warning "${fn}: error on 'jq' parse pods: ${containers}." |
| 464 | return 1 |
| 465 | fi |
| 466 | |
| 467 | [ -n "$kube_cluster_name" ] && echo "$kube_cluster_name" > "$tmp_kube_cluster_name" 2> /dev/null |
| 468 | [ -n "$kube_system_ns" ] && [ -n "$kube_system_uid" ] && echo "$kube_system_uid" > "$tmp_kube_system_ns_uid_file" 2> /dev/null |
| 469 | echo "$containers" > "$tmp_kube_containers_file" 2> /dev/null |
| 470 | fi |
| 471 | |
| 472 | local qos_class |
| 473 | if [[ $clean_id =~ .+(besteffort|burstable) ]]; then |
| 474 | qos_class="${BASH_REMATCH[1]}" |
| 475 | else |
| 476 | qos_class="guaranteed" |
| 477 | fi |
| 478 | |
| 479 | # available labels: |
| 480 | # namespace, pod_name, pod_uid, container_name, container_id, node_name |
| 481 | if [ -n "$cntr_id" ]; then |
| 482 | if [ -n "$labels" ] || labels=$(grep "$cntr_id" <<< "$containers" 2> /dev/null); then |
| 483 | |
| 484 | # --- skip kubevirt helper containers in virt-launcher pods --- |
| 485 | local container_name pod_name |
| 486 | container_name="$(get_lbl_val "$labels" container_name)" |
| 487 | pod_name="$(get_lbl_val "$labels" pod_name)" |
| 488 | |
| 489 | if [[ "$pod_name" == virt-launcher-* ]]; then |
| 490 | case "$container_name" in |
| 491 | volumerootdisk | guest-console-log) |
| 492 | info "${fn}: skipping kubevirt helper container '$container_name' in pod '$pod_name'" |
| 493 | return 3 |
| 494 | ;; |
| 495 | esac |
| 496 | fi |
| 497 | |
| 498 | labels+=',kind="container"' |
| 499 | labels+=",qos_class=\"$qos_class\"" |
| 500 | [ -n "$kube_system_uid" ] && [ "$kube_system_uid" != "null" ] && labels+=",cluster_id=\"$kube_system_uid\"" |
| 501 | [ -n "$kube_cluster_name" ] && [ "$kube_cluster_name" != "unknown" ] && labels+=",cluster_name=\"$kube_cluster_name\"" |
| 502 | name="cntr" |
| 503 | name+="_$(get_lbl_val "$labels" namespace)" |
| 504 | name+="_$(get_lbl_val "$labels" pod_name)" |
| 505 | name+="_$(get_lbl_val "$labels" container_name)" |
| 506 | labels=$(remove_lbl "$labels" "container_id") |
| 507 | labels=$(remove_lbl "$labels" "pod_uid") |
| 508 | labels=$(add_lbl_prefix "$labels" "k8s_") |
| 509 | name+=" $labels" |
| 510 | else |
| 511 | return 2 |
| 512 | fi |
| 513 | elif [ -n "$pod_uid" ]; then |
| 514 | if labels=$(grep "$pod_uid" -m 1 <<< "$containers" 2> /dev/null); then |
| 515 | labels="${labels%%,container_*}" |
| 516 | labels+=',kind="pod"' |
| 517 | labels+=",qos_class=\"$qos_class\"" |
| 518 | [ -n "$kube_system_uid" ] && [ "$kube_system_uid" != "null" ] && labels+=",cluster_id=\"$kube_system_uid\"" |
| 519 | [ -n "$kube_cluster_name" ] && [ "$kube_cluster_name" != "unknown" ] && labels+=",cluster_name=\"$kube_cluster_name\"" |
| 520 | name="pod" |
| 521 | name+="_$(get_lbl_val "$labels" namespace)" |
| 522 | name+="_$(get_lbl_val "$labels" pod_name)" |
| 523 | labels=$(remove_lbl "$labels" "pod_uid") |
| 524 | labels=$(add_lbl_prefix "$labels" "k8s_") |
| 525 | name+=" $labels" |
| 526 | else |
| 527 | return 2 |
| 528 | fi |
| 529 | fi |
| 530 | |
| 531 | # jq filter nonexistent field and nonexistent label value is 'null' |
| 532 | if [[ $name =~ _null(_|$) ]]; then |
| 533 | warning "${fn}: invalid name: $name (cgroup '$id')" |
| 534 | if [ -n "${USE_KUBELET_FOR_PODS_METADATA}" ]; then |
| 535 | # local data is cached and may not contain the correct id |
| 536 | return 2 |
| 537 | fi |
| 538 | return 1 |
| 539 | fi |
| 540 | |
| 541 | echo "$name" |
| 542 | [ -n "$name" ] |
| 543 | return |
| 544 | } |
| 545 | |
| 546 | function k8s_get_name() { |
| 547 | local fn="${FUNCNAME[0]}" |
| 548 | local cgroup_path="${1}" |
| 549 | local id="${2}" |
| 550 | local kubepod_name="" |
| 551 | |
| 552 | kubepod_name=$(k8s_get_kubepod_name "$cgroup_path" "$id") |
| 553 | |
| 554 | case "$?" in |
| 555 | 0) |
| 556 | kubepod_name="k8s_${kubepod_name}" |
| 557 | |
| 558 | local name labels |
| 559 | name=${kubepod_name%% *} |
| 560 | labels=${kubepod_name#* } |
| 561 | |
| 562 | if [ "$name" != "$labels" ]; then |
| 563 | info "${fn}: cgroup '${id}' has chart name '${name}', labels '${labels}" |
| 564 | NAME="$name" |
| 565 | LABELS="$labels" |
| 566 | else |
| 567 | info "${fn}: cgroup '${id}' has chart name '${NAME}'" |
| 568 | NAME="$name" |
| 569 | fi |
| 570 | EXIT_CODE=$EXIT_SUCCESS |
| 571 | ;; |
| 572 | 1) |
| 573 | NAME="k8s_${id}" |
| 574 | warning "${fn}: cannot find the name of cgroup with id '${id}'. Setting name to ${NAME} and enabling it." |
| 575 | EXIT_CODE=$EXIT_SUCCESS |
| 576 | ;; |
| 577 | 2) |
| 578 | NAME="k8s_${id}" |
| 579 | warning "${fn}: cannot find the name of cgroup with id '${id}'. Setting name to ${NAME} and asking for retry." |
| 580 | EXIT_CODE=$EXIT_RETRY |
| 581 | ;; |
| 582 | *) |
| 583 | NAME="k8s_${id}" |
| 584 | warning "${fn}: cannot find the name of cgroup with id '${id}'. Setting name to ${NAME} and disabling it." |
| 585 | EXIT_CODE=$EXIT_DISABLE |
| 586 | ;; |
| 587 | esac |
| 588 | } |
| 589 | |
| 590 | function docker_get_name() { |
| 591 | local id="${1}" |
| 592 | # See https://github.com/netdata/netdata/pull/13523 for details |
| 593 | if command -v snap > /dev/null 2>&1 && snap list docker > /dev/null 2>&1; then |
| 594 | docker_like_get_name_api DOCKER_HOST "${id}" |
| 595 | elif hash docker 2> /dev/null; then |
| 596 | docker_like_get_name_command docker "${id}" |
| 597 | else |
| 598 | docker_like_get_name_api DOCKER_HOST "${id}" || docker_like_get_name_command podman "${id}" |
| 599 | fi |
| 600 | if [ -z "${NAME}" ]; then |
| 601 | warning "cannot find the name of docker container '${id}'" |
| 602 | EXIT_CODE=$EXIT_RETRY |
| 603 | NAME="${id:0:12}" |
| 604 | else |
| 605 | info "docker container '${id}' is named '${NAME}'" |
| 606 | fi |
| 607 | } |
| 608 | |
| 609 | function docker_validate_id() { |
| 610 | local id="${1}" |
| 611 | if [ -n "${id}" ] && { [ ${#id} -eq 64 ] || [ ${#id} -eq 12 ]; }; then |
| 612 | docker_get_name "${id}" |
| 613 | else |
| 614 | error "a docker id cannot be extracted from docker cgroup '${CGROUP}'." |
| 615 | fi |
| 616 | } |
| 617 | |
| 618 | function podman_get_name() { |
| 619 | local id="${1}" |
| 620 | |
| 621 | # for Podman, prefer using the API if we can, as netdata will not normally have access |
| 622 | # to other users' containers, so they will not be visible when running `podman ps` |
| 623 | docker_like_get_name_api PODMAN_HOST "${id}" || docker_like_get_name_command podman "${id}" |
| 624 | |
| 625 | if [ -z "${NAME}" ]; then |
| 626 | warning "cannot find the name of podman container '${id}'" |
| 627 | EXIT_CODE=$EXIT_RETRY |
| 628 | NAME="${id:0:12}" |
| 629 | else |
| 630 | info "podman container '${id}' is named '${NAME}'" |
| 631 | fi |
| 632 | } |
| 633 | |
| 634 | function podman_validate_id() { |
| 635 | local id="${1}" |
| 636 | if [ -n "${id}" ] && [ ${#id} -eq 64 ]; then |
| 637 | podman_get_name "${id}" |
| 638 | else |
| 639 | error "a podman id cannot be extracted from docker cgroup '${CGROUP}'." |
| 640 | fi |
| 641 | } |
| 642 | |
| 643 | # ----------------------------------------------------------------------------- |
| 644 | |
| 645 | DOCKER_HOST="${DOCKER_HOST:=unix:///var/run/docker.sock}" |
| 646 | PODMAN_HOST="${PODMAN_HOST:=unix:///run/podman/podman.sock}" |
| 647 | CGROUP_PATH="${1}" # the path as it is (e.g. '/docker/efcf4c409') |
| 648 | CGROUP="${2//\//_}" # the modified path (e.g. 'docker_efcf4c409') |
| 649 | EXIT_SUCCESS=0 |
| 650 | EXIT_RETRY=2 |
| 651 | EXIT_DISABLE=3 |
| 652 | EXIT_CODE=$EXIT_SUCCESS |
| 653 | NAME= |
| 654 | LABELS= |
| 655 | |
| 656 | # ----------------------------------------------------------------------------- |
| 657 | |
| 658 | if [ -z "${CGROUP}" ]; then |
| 659 | fatal "called without a cgroup name. Nothing to do." |
| 660 | fi |
| 661 | |
| 662 | if [ -z "${NAME}" ]; then |
| 663 | if [[ ${CGROUP} =~ ^.*kubepods.* ]]; then |
| 664 | k8s_get_name "${CGROUP_PATH}" "${CGROUP}" |
| 665 | fi |
| 666 | fi |
| 667 | |
| 668 | if [ -z "${NAME}" ]; then |
| 669 | if [[ ${CGROUP} =~ ^.*docker[-_/\.][a-fA-F0-9]+[-_\.]?.*$ ]]; then |
| 670 | # docker containers |
| 671 | #shellcheck disable=SC1117 |
| 672 | DOCKERID="$(echo "${CGROUP}" | sed "s|^.*docker[-_/]\([a-fA-F0-9]\+\)[-_\.]\?.*$|\1|")" |
| 673 | docker_validate_id "${DOCKERID}" |
| 674 | elif [[ ${CGROUP} =~ ^.*ecs[-_/\.][a-fA-F0-9]+[-_\.]?.*$ ]]; then |
| 675 | # ECS |
| 676 | #shellcheck disable=SC1117 |
| 677 | DOCKERID="$(echo "${CGROUP}" | sed "s|^.*ecs[-_/].*[-_/]\([a-fA-F0-9]\+\)[-_\.]\?.*$|\1|")" |
| 678 | docker_validate_id "${DOCKERID}" |
| 679 | elif [[ ${CGROUP} =~ system.slice_containerd.service_cpuset_[a-fA-F0-9]+[-_\.]?.*$ ]]; then |
| 680 | # docker containers under containerd |
| 681 | #shellcheck disable=SC1117 |
| 682 | DOCKERID="$(echo "${CGROUP}" | sed "s|^.*ystem.slice_containerd.service_cpuset_\([a-fA-F0-9]\+\)[-_\.]\?.*$|\1|")" |
| 683 | docker_validate_id "${DOCKERID}" |
| 684 | elif [[ ${CGROUP} =~ ^.*libpod-[a-fA-F0-9]+.*$ ]]; then |
| 685 | # Podman |
| 686 | PODMANID="$(echo "${CGROUP}" | sed "s|^.*libpod-\(conmon-\)\?\([a-fA-F0-9]\+\).*$|\2|")" |
| 687 | podman_validate_id "${PODMANID}" |
| 688 | |
| 689 | elif [[ ${CGROUP} =~ machine.slice[_/].*\.service ]]; then |
| 690 | # systemd-nspawn |
| 691 | NAME="$(echo "${CGROUP}" | sed 's/.*machine.slice[_\/]\(.*\)\.service/\1/g')" |
| 692 | |
| 693 | elif [[ ${CGROUP} =~ machine.slice_machine.*-lxc ]]; then |
| 694 | # libvirtd / lxc containers |
| 695 | # machine.slice machine-lxc/x2d969/x2dhubud0xians01.scope => lxc/hubud0xians01 |
| 696 | # machine.slice_machine-lxc/x2d969/x2dhubud0xians01.scope/libvirt_init.scope => lxc/hubud0xians01/libvirt_init |
| 697 | NAME="lxc/$(echo "${CGROUP}" | sed 's/machine.slice_machine.*-lxc//; s/[\/_]x2d[[:digit:]]*//; s/[\/_]x2d//g; s/\.scope//g')" |
| 698 | elif [[ ${CGROUP} =~ machine.slice_machine.*-qemu ]]; then |
| 699 | # libvirtd / qemu virtual machines |
| 700 | # machine.slice_machine-qemu_x2d1_x2dopnsense.scope => qemu_opnsense |
| 701 | NAME="qemu_$(echo "${CGROUP}" | sed 's/machine.slice_machine.*-qemu//; s/[\/_]x2d[[:digit:]]*//; s/[\/_]x2d//g; s/\.scope//g')" |
| 702 | |
| 703 | elif [[ ${CGROUP} =~ machine_.*\.libvirt-qemu ]]; then |
| 704 | # libvirtd / qemu virtual machines |
| 705 | NAME="qemu_$(echo "${CGROUP}" | sed 's/^machine_//; s/\.libvirt-qemu$//; s/-/_/;')" |
| 706 | |
| 707 | elif [[ ${CGROUP} =~ qemu.slice_([0-9]+).scope && -d "${NETDATA_HOST_PREFIX}/etc/pve" ]]; then |
| 708 | # Proxmox VMs |
| 709 | FILENAME="${NETDATA_HOST_PREFIX}/etc/pve/qemu-server/${BASH_REMATCH[1]}.conf" |
| 710 | if [[ -f $FILENAME && -r $FILENAME ]]; then |
| 711 | NAME="qemu_$(grep -e '^name: ' "${FILENAME}" | head -1 | sed -rn 's|\s*name\s*:\s*(.*)?$|\1|p')" |
| 712 | else |
| 713 | error "proxmox config file missing ${FILENAME} or netdata does not have read access. Please ensure netdata is a member of www-data group." |
| 714 | fi |
| 715 | elif [[ ${CGROUP} =~ lxc_([0-9]+) && -d "${NETDATA_HOST_PREFIX}/etc/pve" ]]; then |
| 716 | # Proxmox Containers (LXC) |
| 717 | FILENAME="${NETDATA_HOST_PREFIX}/etc/pve/lxc/${BASH_REMATCH[1]}.conf" |
| 718 | if [[ -f ${FILENAME} && -r ${FILENAME} ]]; then |
| 719 | NAME=$(grep -e '^hostname: ' "${FILENAME}" | head -1 | sed -rn 's|\s*hostname\s*:\s*(.*)?$|\1|p') |
| 720 | else |
| 721 | error "proxmox config file missing ${FILENAME} or netdata does not have read access. Please ensure netdata is a member of www-data group." |
| 722 | fi |
| 723 | elif [[ ${CGROUP} =~ lxc.payload.* ]]; then |
| 724 | # LXC 4.0 |
| 725 | NAME="$(echo "${CGROUP}" | sed 's/lxc\.payload\.\(.*\)/\1/g')" |
| 726 | fi |
| 727 | |
| 728 | [ -z "${NAME}" ] && NAME="${CGROUP}" |
| 729 | [ ${#NAME} -gt 100 ] && NAME="${NAME:0:100}" |
| 730 | fi |
| 731 | |
| 732 | NAME="${NAME// /_}" |
| 733 | |
| 734 | info "cgroup '${CGROUP}' is called '${NAME}', labels '${LABELS}'" |
| 735 | if [ -n "$LABELS" ]; then |
| 736 | echo "${NAME} ${LABELS}" |
| 737 | else |
| 738 | echo "${NAME}" |
| 739 | fi |
| 740 | |
| 741 | exit ${EXIT_CODE} |