[elasticsearch] Fix issue with readinessProbe causing outages (#638)

elastic · May 28, 2020 · 5cb6c06 · 5cb6c06
1 parent 669e799
commit 5cb6c06
Show file tree

Hide file tree

Showing 2 changed files with 199 additions and 10 deletions.
diff --git a/BREAKING_CHANGES.md b/BREAKING_CHANGES.md
@@ -0,0 +1,179 @@
+# Breaking changes
+<!-- START doctoc generated TOC please keep comment here to allow auto update -->
+<!-- DON'T EDIT THIS SECTION, INSTEAD RE-RUN doctoc TO UPDATE -->
+
+
+- [7.7.0 - 2020/05/13](#770---20200513)
+  - [Known Issues](#known-issues)
+  - [GA support](#ga-support)
+  - [New branching model](#new-branching-model)
+  - [Filebeat container inputs](#filebeat-container-inputs)
+  - [Metricbeat upgrade issue](#metricbeat-upgrade-issue)
+  - [Metricbeat split values for daemonset and deployment](#metricbeat-split-values-for-daemonset-and-deployment)
+- [6.8.9 - 2020/05/13](#689---20200513)
+- [7.6.2 - 2020/03/31](#762---20200331)
+  - [Kibana default resources](#kibana-default-resources)
+- [7.6.0 - 2020/02/11](#760---20200211)
+  - [Elasticsearch default resources](#elasticsearch-default-resources)
+- [7.5.0 - 2019/12/02](#750---20191202)
+  - [Metricbeat kube-state-metrics upgrade](#metricbeat-kube-state-metrics-upgrade)
+- [7.0.0-alpha1 - 2019/04/17](#700-alpha1---20190417)
+  - [Elasticsearch upgrade from 6.x](#elasticsearch-upgrade-from-6x)
+
+<!-- END doctoc generated TOC please keep comment here to allow auto update -->
+<!-- Use this to update TOC: -->
+<!-- docker run --rm -it -v $(pwd):/usr/src jorgeandrada/doctoc --github -->
+
+
+## 7.7.0 - 2020/05/13
+
+### Known Issues
+
+Elasticsearch nodes could be restarted too quickly during an upgrade or rolling restart, potentially resulting in service disruption.
+This is due to a bug introduced by the changes to the Elasticsearch `readinessProbe` in [#586][].
+
+### GA support
+
+Elasticsearch, Kibana, Filebeat and Metricbeat are moving from beta to GA and
+are supported by Elastic following these limitations:
+- only released charts coming from [Elastic Helm repo][] or
+[GitHub releases][] are supported.
+- released charts are only supported when using the same chart version and
+application version (ie: using 7.7.0 chart with 6.8.8 or 7.6.2 application is
+not supported).
+
+### New branching model
+
+Elastic Helm charts repository is now following a [new branching model][]:
+- `master` branch is now a development branch for next major release.
+- new `7.x` branch is a development branch for next minor release using SNAPSHOT
+Docker images.
+- new `7.7` branch is a development branch for next patch release using SNAPSHOT
+Docker images
+
+### Filebeat container inputs
+
+Filebeat chart default config is now using [container input][] instead of
+[docker input][] in [#568][].
+
+### Metricbeat upgrade issue
+
+Metricbeat upgrade are failing with
+`spec.selector: Invalid value: ... field is immutable` error. This is related to
+Metricbeat deployment selector including chart version which is not immutable.
+You should use `helm upgrade --force` to upgrade Metricbeat. See [#621][] for
+more details.
+
+### Metricbeat split values for daemonset and deployment
+
+Metricbeat is now using dedicated values for daemonset and deployment config.
+The old values are still working but are now deprecated. See [#572][] for more
+details.
+
+## 6.8.9 - 2020/05/13
+
+See [7.7.0 Breaking changes](#770---20200513)
+
+
+## 7.6.2 - 2020/03/31
+
+### Kibana default resources
+
+Kibana default resources (cpu/memory requests and limits) are increased in
+[#540][].
+
+This change may impact cpu/memory available resources capacity in your
+Kubernetes cluster.
+
+To come back to former default values, use the following values:
+
+```yaml
+extraEnvs:
+- name: "NODE_OPTIONS"
+  value: ""
+resources:
+  requests:
+    cpu: "100m"
+    memory: "500Mi"
+  limits:
+    cpu: "1000m"
+    memory: "1Gi"
+```
+
+
+## 7.6.0 - 2020/02/11
+
+### Elasticsearch default resources
+
+Elasticsearch default cpu requests is increased in [#458][] following our
+recommendation that resources requests and limits should have the same values.
+
+This change may impact available cpu capacity in your Kubernetes cluster.
+
+To come back to former default values, use the following values:
+
+```yaml
+resources:
+  requests:
+    cpu: "100m"
+```
+
+
+## 7.5.0 - 2019/12/02
+
+### Metricbeat kube-state-metrics upgrade
+
+[kube-state-metrics][] chart dependency is upgraded from 1.6.0 to 2.4.1 in
+[#352][]. This is causing Metricbeat chart upgrade from versions < 7.5.0 failing
+with the following error:
+
+```
+UPGRADE FAILED
+Error: Deployment.apps "metricbeat-kube-state-metrics" is invalid: spec.selector: Invalid value: v1.LabelSelector{MatchLabels:map[string]string{"app.kubernetes.io/name":"kube-state-metrics"}, MatchExpressions:[]v1.LabelSelectorRequirement(nil)}: field is immutable && Deployment.apps "metricbeat-metricbeat-metrics" is invalid: spec.selector: Invalid value: v1.LabelSelector{MatchLabels:map[string]string{"app":"metricbeat-metricbeat-metrics", "chart":"metricbeat-7.5.0", "heritage":"Tiller", "release":"metricbeat"}, MatchExpressions:[]v1.LabelSelectorRequirement(nil)}: field is immutable
+Error: UPGRADE FAILED: Deployment.apps "metricbeat-kube-state-metrics" is invalid: spec.selector: Invalid value: v1.LabelSelector{MatchLabels:map[string]string{"app.kubernetes.io/name":"kube-state-metrics"}, MatchExpressions:[]v1.LabelSelectorRequirement(nil)}: field is immutable && Deployment.apps "metricbeat-metricbeat-metrics" is invalid: spec.selector: Invalid value: v1.LabelSelector{MatchLabels:map[string]string{"app":"metricbeat-metricbeat-metrics", "chart":"metricbeat-7.5.0", "heritage":"Tiller", "release":"metricbeat"}, MatchExpressions:[]v1.LabelSelectorRequirement(nil)}: field is immutable
+```
+
+The workaround is to use `--force` argument for `helm upgrade` command which
+will force Metricbeat resources update through delete/recreate.
+
+
+## 7.0.0-alpha1 - 2019/04/17
+
+### Elasticsearch upgrade from 6.x
+
+If you were using the default Elasticsearch version from the previous release
+(6.6.2-alpha1) you will first need to upgrade to Elasticsearch 6.7.1 before
+being able to upgrade to 7.0.0. You can do this by adding this to your values
+file:
+
+```yaml
+esMajorVersion: 6
+imageTag: 6.7.1
+```
+
+If you are upgrading an existing cluster that did not override the default
+`storageClassName` you will now need to specify the `storageClassName`. This
+only affects existing clusters and was changed in [#94][]. The advantage of this
+is that now the Helm chart will just use the default `storageClassName` rather
+than needing to override it for any providers where it is not called `standard`.
+
+```
+volumeClaimTemplate:
+  storageClassName: "standard"
+```
+
+
+[#94]: https://github.com/elastic/helm-charts/pull/94
+[#352]: https://github.com/elastic/helm-charts/pull/352
+[#458]: https://github.com/elastic/helm-charts/pull/458
+[#540]: https://github.com/elastic/helm-charts/pull/540
+[#568]: https://github.com/elastic/helm-charts/pull/568
+[#572]: https://github.com/elastic/helm-charts/pull/572
+[#586]: https://github.com/elastic/helm-charts/pull/586
+[#621]: https://github.com/elastic/helm-charts/pull/621
+[container input]: https://www.elastic.co/guide/en/beats/filebeat/7.7/filebeat-input-container.html
+[docker input]: https://www.elastic.co/guide/en/beats/filebeat/7.7/filebeat-input-docker.html
+[elastic helm repo]: https://helm.elastic.co
+[github releases]: https://github.com/elastic/helm-charts/releases
+[new branching model]: https://github.com/elastic/helm-charts/blob/master/CONTRIBUTING.md#branching
+[kube-state-metrics]: https://github.com/helm/charts/tree/master/stable/kube-state-metrics
diff --git a/elasticsearch/templates/statefulset.yaml b/elasticsearch/templates/statefulset.yaml
@@ -213,22 +213,32 @@ spec:
               - -c
               - |
                 #!/usr/bin/env bash -e
-                # If the node is starting up wait for the cluster to be ready (request params: '{{ .Values.clusterHealthCheckParams }}' )
+                # If the node is starting up wait for the cluster to be ready (request params: "{{ .Values.clusterHealthCheckParams }}" )
                 # Once it has started only check that the node itself is responding
                 START_FILE=/tmp/.es_start_file
 
-                if [ -n "${ELASTIC_USERNAME}" ] && [ -n "${ELASTIC_PASSWORD}" ]; then
-                  BASIC_AUTH="-u ${ELASTIC_USERNAME}:${ELASTIC_PASSWORD}"
-                else
-                  BASIC_AUTH=''
-                fi
+                http () {
+                  local path="${1}"
+                  local args="${2}"
+                  set -- -XGET -s
+
+                  if [ "$args" != "" ]; then
+                    set -- "$@" $args
+                  fi
+
+                  if [ -n "${ELASTIC_USERNAME}" ] && [ -n "${ELASTIC_PASSWORD}" ]; then
+                    set -- "$@" -u "${ELASTIC_USERNAME}:${ELASTIC_PASSWORD}"
+                  fi
+
+                  curl --output /dev/null -k "$@" "{{ .Values.protocol }}://127.0.0.1:{{ .Values.httpPort }}${path}"
+                }
 
                 if [ -f "${START_FILE}" ]; then
                   echo 'Elasticsearch is already running, lets check the node is healthy'
-                  HTTP_CODE=$(curl -XGET -s -k ${BASIC_AUTH} -o /dev/null -w '%{http_code}' {{ .Values.protocol }}://127.0.0.1:{{ .Values.httpPort }}/)
+                  HTTP_CODE=$(http "/" "-w %{http_code}")
                   RC=$?
                   if [[ ${RC} -ne 0 ]]; then
-                    echo "curl -XGET -s -k \${BASIC_AUTH} -o /dev/null -w '%{http_code}' {{ .Values.protocol }}://127.0.0.1:{{ .Values.httpPort }}/ failed with RC ${RC}"
+                    echo "curl --output /dev/null -k -XGET -s -w '%{http_code}' \${BASIC_AUTH} {{ .Values.protocol }}://127.0.0.1:{{ .Values.httpPort }}/ failed with RC ${RC}"
                     exit ${RC}
                   fi
                   # ready if HTTP code 200, 503 is tolerable if ES version is 6.x
@@ -237,13 +247,13 @@ spec:
                   elif [[ ${HTTP_CODE} == "503" && "{{ include "elasticsearch.esMajorVersion" . }}" == "6" ]]; then
                     exit 0
                   else
-                    echo "curl -XGET -s -k \${BASIC_AUTH} -o /dev/null -w '%{http_code}' {{ .Values.protocol }}://127.0.0.1:{{ .Values.httpPort }}/ failed with HTTP code ${HTTP_CODE}"
+                    echo "curl --output /dev/null -k -XGET -s -w '%{http_code}' \${BASIC_AUTH} {{ .Values.protocol }}://127.0.0.1:{{ .Values.httpPort }}/ failed with HTTP code ${HTTP_CODE}"
                     exit 1
                   fi
 
                 else
                   echo 'Waiting for elasticsearch cluster to become ready (request params: "{{ .Values.clusterHealthCheckParams }}" )'
-                  if curl -XGET -s -k --fail ${BASIC_AUTH} {{ .Values.protocol }}://127.0.0.1:{{ .Values.httpPort }}/_cluster/health?{{ .Values.clusterHealthCheckParams }} ; then
+                  if http "/_cluster/health?{{ .Values.clusterHealthCheckParams }}" "--fail" ; then
                     touch ${START_FILE}
                     exit 0
                   else