cgroup-name.sh.in 8.6 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272
  1. #!/usr/bin/env bash
  2. #shellcheck disable=SC2001
  3. # netdata
  4. # real-time performance and health monitoring, done right!
  5. # (C) 2016 Costa Tsaousis <costa@tsaousis.gr>
  6. # SPDX-License-Identifier: GPL-3.0-or-later
  7. #
  8. # Script to find a better name for cgroups
  9. #
  10. export PATH="${PATH}:/sbin:/usr/sbin:/usr/local/sbin"
  11. export LC_ALL=C
  12. # -----------------------------------------------------------------------------
  13. PROGRAM_NAME="$(basename "${0}")"
  14. logdate() {
  15. date "+%Y-%m-%d %H:%M:%S"
  16. }
  17. log() {
  18. local status="${1}"
  19. shift
  20. echo >&2 "$(logdate): ${PROGRAM_NAME}: ${status}: ${*}"
  21. }
  22. warning() {
  23. log WARNING "${@}"
  24. }
  25. error() {
  26. log ERROR "${@}"
  27. }
  28. info() {
  29. log INFO "${@}"
  30. }
  31. fatal() {
  32. log FATAL "${@}"
  33. exit 1
  34. }
  35. function docker_like_get_name_command() {
  36. local command="${1}"
  37. local id="${2}"
  38. info "Running command: ${command} ps --filter=id=\"${id}\" --format=\"{{.Names}}\""
  39. NAME="$(${command} ps --filter=id="${id}" --format="{{.Names}}")"
  40. return 0
  41. }
  42. function docker_like_get_name_api() {
  43. local host_var="${1}"
  44. local host="${!host_var}"
  45. local path="/containers/${2}/json"
  46. if [ -z "${host}" ]; then
  47. warning "No ${host_var} is set"
  48. return 1
  49. fi
  50. if ! command -v jq >/dev/null 2>&1; then
  51. warning "Can't find jq command line tool. jq is required for netdata to retrieve container name using ${host} API, falling back to docker ps"
  52. return 1
  53. fi
  54. if [ -S "${host}" ]; then
  55. info "Running API command: curl --unix-socket \"${host}\" http://localhost${path}"
  56. JSON=$(curl -sS --unix-socket "${host}" "http://localhost${path}")
  57. else
  58. info "Running API command: curl \"${host}${path}\""
  59. JSON=$(curl -sS "${host}${path}")
  60. fi
  61. NAME=$(echo "${JSON}" | jq -r .Name,.Config.Hostname | grep -v null | head -n1 | sed 's|^/||')
  62. return 0
  63. }
  64. function k8s_get_name() {
  65. if [[ "${1}" =~ ^kube.*_pod.*$ ]]; then
  66. local id="${1##*_pod}"
  67. else
  68. # Take the last part of the delimited path identifier (expecting either _ or / as a delimiter).
  69. local id="${1##*_}"
  70. if [ "${id}" == "${1}" ]; then
  71. id="${1##*/}"
  72. fi
  73. fi
  74. if command -v jq >/dev/null 2>&1; then
  75. if [ -n "${KUBERNETES_SERVICE_HOST}" ] && [ -n "${KUBERNETES_PORT_443_TCP_PORT}" ]; then
  76. KUBE_TOKEN="$(</var/run/secrets/kubernetes.io/serviceaccount/token)"
  77. NAME="$(
  78. curl -sSk -H "Authorization: Bearer $KUBE_TOKEN" "https://$KUBERNETES_SERVICE_HOST:$KUBERNETES_PORT_443_TCP_PORT/api/v1/pods" |
  79. jq -r '.items[] | "k8s_\(.metadata.namespace)_\(.metadata.name)_\(.metadata.uid)_" + (.status.containerStatuses[]? | "\(.name) \(.containerID)")' |
  80. grep "$id" |
  81. cut -d' ' -f1
  82. )"
  83. elif ps -C kubelet >/dev/null 2>&1 && command -v kubectl >/dev/null 2>&1; then
  84. if [[ -z ${KUBE_CONFIG+x} ]]; then
  85. KUBE_CONFIG="/etc/kubernetes/admin.conf"
  86. fi
  87. if kubectl --kubeconfig=$KUBE_CONFIG get pod --all-namespaces >/dev/null 2>&1; then
  88. #shellcheck disable=SC2086
  89. NAME="$(kubectl --kubeconfig=$KUBE_CONFIG get pod --all-namespaces --output='json' |
  90. jq -r '.items[] | "k8s_\(.metadata.namespace)_\(.metadata.name)_\(.metadata.uid)_" + (.status.containerStatuses[]? | "\(.name) \(.containerID)")' |
  91. grep "$id" |
  92. cut -d' ' -f1
  93. )"
  94. else
  95. warning "kubectl cannot get pod list, check for configuration file in $KUBE_CONFIG, or set this path to env \$KUBE_CONFIG"
  96. fi
  97. fi
  98. else
  99. warning "jq command not available, k8s_get_name() cannot execute. Please install jq should you wish for k8s to be fully functional"
  100. fi
  101. if [ -z "${NAME}" ]; then
  102. warning "cannot find the name of k8s pod with containerID '${id}'. Setting name to ${id} and disabling it"
  103. NAME="${id}"
  104. NAME_NOT_FOUND=3
  105. else
  106. info "k8s containerID '${id}' has chart name (namespace_podname_poduid_containername) '${NAME}'"
  107. fi
  108. }
  109. function docker_get_name() {
  110. local id="${1}"
  111. if hash docker 2>/dev/null; then
  112. docker_like_get_name_command docker "${id}"
  113. else
  114. docker_like_get_name_api DOCKER_HOST "${id}" || docker_like_get_name_command podman "${id}"
  115. fi
  116. if [ -z "${NAME}" ]; then
  117. warning "cannot find the name of docker container '${id}'"
  118. NAME_NOT_FOUND=2
  119. NAME="${id:0:12}"
  120. else
  121. info "docker container '${id}' is named '${NAME}'"
  122. fi
  123. }
  124. function docker_validate_id() {
  125. local id="${1}"
  126. if [ -n "${id}" ] && { [ ${#id} -eq 64 ] || [ ${#id} -eq 12 ]; }; then
  127. docker_get_name "${id}"
  128. else
  129. error "a docker id cannot be extracted from docker cgroup '${CGROUP}'."
  130. fi
  131. }
  132. function podman_get_name() {
  133. local id="${1}"
  134. # for Podman, prefer using the API if we can, as netdata will not normally have access
  135. # to other users' containers, so they will not be visible when running `podman ps`
  136. docker_like_get_name_api PODMAN_HOST "${id}" || docker_like_get_name_command podman "${id}"
  137. if [ -z "${NAME}" ]; then
  138. warning "cannot find the name of podman container '${id}'"
  139. NAME_NOT_FOUND=2
  140. NAME="${id:0:12}"
  141. else
  142. info "podman container '${id}' is named '${NAME}'"
  143. fi
  144. }
  145. function podman_validate_id() {
  146. local id="${1}"
  147. if [ -n "${id}" ] && [ ${#id} -eq 64 ]; then
  148. podman_get_name "${id}"
  149. else
  150. error "a podman id cannot be extracted from docker cgroup '${CGROUP}'."
  151. fi
  152. }
  153. # -----------------------------------------------------------------------------
  154. [ -z "${NETDATA_USER_CONFIG_DIR}" ] && NETDATA_USER_CONFIG_DIR="@configdir_POST@"
  155. [ -z "${NETDATA_STOCK_CONFIG_DIR}" ] && NETDATA_STOCK_CONFIG_DIR="@libconfigdir_POST@"
  156. DOCKER_HOST="${DOCKER_HOST:=/var/run/docker.sock}"
  157. PODMAN_HOST="${PODMAN_HOST:=/run/podman/podman.sock}"
  158. CGROUP="${1}"
  159. NAME_NOT_FOUND=0
  160. NAME=
  161. # -----------------------------------------------------------------------------
  162. if [ -z "${CGROUP}" ]; then
  163. fatal "called without a cgroup name. Nothing to do."
  164. fi
  165. for CONFIG in "${NETDATA_USER_CONFIG_DIR}/cgroups-names.conf" "${NETDATA_STOCK_CONFIG_DIR}/cgroups-names.conf"; do
  166. if [ -f "${CONFIG}" ]; then
  167. NAME="$(grep "^${CGROUP} " "${CONFIG}" | sed 's/[[:space:]]\+/ /g' | cut -d ' ' -f 2)"
  168. if [ -z "${NAME}" ]; then
  169. info "cannot find cgroup '${CGROUP}' in '${CONFIG}'."
  170. else
  171. break
  172. fi
  173. #else
  174. # info "configuration file '${CONFIG}' is not available."
  175. fi
  176. done
  177. if [ -z "${NAME}" ]; then
  178. if [[ ${CGROUP} =~ ^.*kubepods.* ]]; then
  179. k8s_get_name "${CGROUP}"
  180. fi
  181. fi
  182. if [ -z "${NAME}" ]; then
  183. if [[ ${CGROUP} =~ ^.*docker[-_/\.][a-fA-F0-9]+[-_\.]?.*$ ]]; then
  184. # docker containers
  185. #shellcheck disable=SC1117
  186. DOCKERID="$(echo "${CGROUP}" | sed "s|^.*docker[-_/]\([a-fA-F0-9]\+\)[-_\.]\?.*$|\1|")"
  187. docker_validate_id "${DOCKERID}"
  188. elif [[ ${CGROUP} =~ ^.*ecs[-_/\.][a-fA-F0-9]+[-_\.]?.*$ ]]; then
  189. # ECS
  190. #shellcheck disable=SC1117
  191. DOCKERID="$(echo "${CGROUP}" | sed "s|^.*ecs[-_/].*[-_/]\([a-fA-F0-9]\+\)[-_\.]\?.*$|\1|")"
  192. docker_validate_id "${DOCKERID}"
  193. elif [[ ${CGROUP} =~ ^.*libpod-[a-fA-F0-9]+.*$ ]]; then
  194. # Podman
  195. PODMANID="$(echo "${CGROUP}" | sed "s|^.*libpod-\([a-fA-F0-9]\+\).*$|\1|")"
  196. podman_validate_id "${PODMANID}"
  197. elif [[ ${CGROUP} =~ machine.slice[_/].*\.service ]]; then
  198. # systemd-nspawn
  199. NAME="$(echo "${CGROUP}" | sed 's/.*machine.slice[_\/]\(.*\)\.service/\1/g')"
  200. elif [[ ${CGROUP} =~ machine.slice_machine.*-qemu ]]; then
  201. # libvirtd / qemu virtual machines
  202. # NAME="$(echo ${CGROUP} | sed 's/machine.slice_machine.*-qemu//; s/\/x2d//; s/\/x2d/\-/g; s/\.scope//g')"
  203. NAME="qemu_$(echo "${CGROUP}" | sed 's/machine.slice_machine.*-qemu//; s/\/x2d[[:digit:]]*//; s/\/x2d//g; s/\.scope//g')"
  204. elif [[ ${CGROUP} =~ machine_.*\.libvirt-qemu ]]; then
  205. # libvirtd / qemu virtual machines
  206. NAME="qemu_$(echo "${CGROUP}" | sed 's/^machine_//; s/\.libvirt-qemu$//; s/-/_/;')"
  207. elif [[ ${CGROUP} =~ qemu.slice_([0-9]+).scope && -d /etc/pve ]]; then
  208. # Proxmox VMs
  209. FILENAME="/etc/pve/qemu-server/${BASH_REMATCH[1]}.conf"
  210. if [[ -f $FILENAME && -r $FILENAME ]]; then
  211. NAME="qemu_$(grep -e '^name: ' "/etc/pve/qemu-server/${BASH_REMATCH[1]}.conf" | head -1 | sed -rn 's|\s*name\s*:\s*(.*)?$|\1|p')"
  212. else
  213. error "proxmox config file missing ${FILENAME} or netdata does not have read access. Please ensure netdata is a member of www-data group."
  214. fi
  215. elif [[ ${CGROUP} =~ lxc_([0-9]+) && -d /etc/pve ]]; then
  216. # Proxmox Containers (LXC)
  217. FILENAME="/etc/pve/lxc/${BASH_REMATCH[1]}.conf"
  218. if [[ -f ${FILENAME} && -r ${FILENAME} ]]; then
  219. NAME=$(grep -e '^hostname: ' "/etc/pve/lxc/${BASH_REMATCH[1]}.conf" | head -1 | sed -rn 's|\s*hostname\s*:\s*(.*)?$|\1|p')
  220. else
  221. error "proxmox config file missing ${FILENAME} or netdata does not have read access. Please ensure netdata is a member of www-data group."
  222. fi
  223. elif [[ ${CGROUP} =~ lxc.payload.* ]]; then
  224. # LXC 4.0
  225. NAME="$(echo "${CGROUP}" | sed 's/lxc\.payload\.\(.*\)/\1/g')"
  226. fi
  227. [ -z "${NAME}" ] && NAME="${CGROUP}"
  228. [ ${#NAME} -gt 100 ] && NAME="${NAME:0:100}"
  229. fi
  230. info "cgroup '${CGROUP}' is called '${NAME}'"
  231. echo "${NAME}"
  232. exit ${NAME_NOT_FOUND}