| garciadeblas | cf603f5 | 2025-06-04 11:57:28 +0200 | [diff] [blame] | 1 | #!/usr/bin/env bash |
| 2 | |
| 3 | # Copyright 2020 Telefónica Investigación y Desarrollo S.A.U. |
| 4 | # |
| 5 | # Licensed under the Apache License, Version 2.0 (the "License"); |
| 6 | # you may not use this file except in compliance with the License. |
| 7 | # You may obtain a copy of the License at |
| 8 | # |
| 9 | # http://www.apache.org/licenses/LICENSE-2.0 |
| 10 | # |
| 11 | # Unless required by applicable law or agreed to in writing, software |
| 12 | # distributed under the License is distributed on an "AS IS" BASIS, |
| 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 14 | # See the License for the specific language governing permissions and |
| 15 | # limitations under the License. |
| 16 | |
| 17 | HERE=$(dirname "$(readlink --canonicalize "$BASH_SOURCE")") |
| 18 | |
| 19 | source "${HERE}/00-default-install-options.rc" |
| 20 | [ ! -f "${OSM_HOME_DIR}/user-install-options.rc" ] || source "${OSM_HOME_DIR}/user-install-options.rc" |
| 21 | source "${CREDENTIALS_DIR}/git_environment.rc" |
| 22 | export KUBECONFIG="${OSM_HOME_DIR}/clusters/kubeconfig-osm.yaml" |
| 23 | |
| 24 | # Default values |
| 25 | sampling_period=5 # seconds |
| garciadeblas | b59b320 | 2025-06-19 10:09:16 +0200 | [diff] [blame] | 26 | time_for_readiness=2 # minutes ready |
| garciadeblas | cf603f5 | 2025-06-04 11:57:28 +0200 | [diff] [blame] | 27 | time_for_failure=7 # minutes broken |
| 28 | |
| 29 | oks_threshold=$((time_for_readiness*60/${sampling_period})) # No. ok samples to declare the system ready |
| 30 | failures_threshold=$((time_for_failure*60/${sampling_period})) # No. nok samples to declare the system broken |
| 31 | failures_in_a_row=0 |
| 32 | oks_in_a_row=0 |
| 33 | |
| 34 | |
| 35 | # Show status of the OSM services deployed with helm |
| 36 | echo "helm -n ${OSM_NAMESPACE} list" |
| 37 | helm -n ${OSM_NAMESPACE} list |
| 38 | echo "helm -n ${OSM_NAMESPACE} status ${OSM_HELM_RELEASE}" |
| 39 | helm -n ${OSM_NAMESPACE} status ${OSM_HELM_RELEASE} |
| 40 | |
| 41 | #################################################################################### |
| 42 | # Loop to check system readiness |
| 43 | #################################################################################### |
| 44 | while [[ (${failures_in_a_row} -lt ${failures_threshold}) && (${oks_in_a_row} -lt ${oks_threshold}) ]] |
| 45 | do |
| 46 | |
| 47 | # State of Deployments |
| 48 | DEPLOYMENTS_STATE=$(kubectl get deployment -n ${OSM_NAMESPACE} --no-headers 2>&1) |
| 49 | DEPLOYMENTS_READY=$(echo "${DEPLOYMENTS_STATE}" | awk '$2=="1/1" && $4=="1" {printf ("%20s\t%s\t%s\n", $1, $2, $4)}') |
| 50 | DEPLOYMENTS_NOT_READY=$(echo "${DEPLOYMENTS_STATE}" | awk '$2!="1/1" || $4!="1" {printf ("%20s\t%s\t%s\n", $1, $2, $4)}') |
| 51 | COUNT_DEPLOYMENTS_READY=$(echo "${DEPLOYMENTS_READY}" | grep -v -e '^$' | wc -l || true) |
| 52 | COUNT_DEPLOYMENTS_NOT_READY=$(echo "${DEPLOYMENTS_NOT_READY}" | grep -v -e '^$' | wc -l || true) |
| 53 | |
| 54 | # State of Statefulsets |
| 55 | STS_STATE=$(kubectl get statefulset -n ${OSM_NAMESPACE} --no-headers 2>&1) |
| 56 | STS_READY=$(echo "${STS_STATE}" | awk '$2=="1/1" || $2=="2/2" || $2=="3/3" {printf ("%20s\t%s\t%s\n", $1, $2, $4)}') |
| 57 | STS_NOT_READY=$(echo "${STS_STATE}" | awk '$2!="1/1" && $2!="2/2" && $2!="3/3" {printf ("%20s\t%s\t%s\n", $1, $2, $4)}') |
| 58 | COUNT_STS_READY=$(echo "${STS_READY}" | grep -v -e '^$' | wc -l || true) |
| 59 | COUNT_STS_NOT_READY=$(echo "${STS_NOT_READY}" | grep -v -e '^$' | wc -l || true) |
| 60 | |
| 61 | # OK sample |
| 62 | if [[ $((${COUNT_DEPLOYMENTS_NOT_READY}+${COUNT_STS_NOT_READY})) -eq 0 ]] |
| 63 | then |
| 64 | ((++oks_in_a_row)) |
| 65 | failures_in_a_row=0 |
| 66 | echo -ne ===\> Successful checks: "${oks_in_a_row}"/${oks_threshold}\\r |
| 67 | # NOK sample |
| 68 | else |
| 69 | ((++failures_in_a_row)) |
| 70 | oks_in_a_row=0 |
| 71 | echo |
| 72 | echo Bootstraping... "${failures_in_a_row}" attempts of ${failures_threshold} |
| 73 | |
| 74 | # Reports failed deployments |
| 75 | if [[ "${COUNT_DEPLOYMENTS_NOT_READY}" -ne 0 ]] |
| 76 | then |
| 77 | echo ${COUNT_DEPLOYMENTS_NOT_READY} of $((${COUNT_DEPLOYMENTS_NOT_READY}+${COUNT_DEPLOYMENTS_READY})) deployments starting: |
| 78 | echo "${DEPLOYMENTS_NOT_READY}" |
| 79 | echo |
| 80 | fi |
| 81 | |
| 82 | # Reports failed statefulsets |
| 83 | if [[ "${COUNT_STS_NOT_READY}" -ne 0 ]] |
| 84 | then |
| 85 | echo ${COUNT_STS_NOT_READY} of $((${COUNT_STS_NOT_READY}+${COUNT_STS_READY})) statefulsets starting: |
| 86 | echo "${STS_NOT_READY}" |
| 87 | echo |
| 88 | fi |
| 89 | fi |
| 90 | |
| 91 | #------------ NEXT SAMPLE |
| 92 | sleep ${sampling_period} |
| 93 | |
| 94 | done |
| 95 | |
| 96 | |
| 97 | #################################################################################### |
| 98 | # OUTCOME |
| 99 | #################################################################################### |
| 100 | if [[ (${failures_in_a_row} -ge ${failures_threshold}) ]] |
| 101 | then |
| 102 | echo |
| 103 | echo SYSTEM IS BROKEN |
| 104 | exit 1 |
| 105 | else |
| 106 | echo |
| 107 | echo SYSTEM IS READY |
| 108 | fi |
| 109 | |
| 110 | exit 0 |