Solr Operator Upgrade Notes
Please carefully read the entries for all versions between the version you are running and the version you want to upgrade to.
Ensure to read the Upgrade Warnings and Notes for the version you are upgrading to as well as the versions you are skipping.
If you want to skip versions when upgrading, be sure to check out the [upgrading-minor-versions-vxa-vxb] and [upgrading-patch-versions-vxya-vzyb] sections.
Version Compatibility Matrixes
Kubernetes Versions
| Solr Operator Version | 1.15 |
1.16 - 1.18 |
1.19 - 1.20 |
1.21 |
1.22+ |
|---|---|---|---|---|---|
|
YES |
YES |
YES |
YES |
NO |
|
NO |
YES |
YES |
YES |
NO |
|
NO |
YES |
YES |
YES |
NO |
|
NO |
YES |
YES |
YES |
NO |
|
NO |
YES |
YES |
YES |
NO |
|
NO |
NO |
YES |
YES |
YES |
|
NO |
NO |
YES |
YES |
YES |
|
NO |
NO |
NO |
YES |
YES |
|
NO |
NO |
NO |
NO |
YES |
|
NO |
NO |
NO |
NO |
YES |
|
NO |
NO |
NO |
NO |
YES |
Solr Versions
| Solr Operator Version | 7.7 |
8.0 - 8.10 |
8.11 |
9.0 - 9.3 |
9.4+ |
10.0+ |
|---|---|---|---|---|---|---|
|
YES |
YES |
YES |
NO |
NO |
NO |
|
YES |
YES |
YES |
NO |
NO |
NO |
|
YES |
YES |
YES |
NO |
NO |
NO |
|
YES |
YES |
YES |
NO |
NO |
NO |
|
YES |
YES |
YES |
NO |
NO |
NO |
|
YES |
YES |
YES |
YES 1 |
YES 1 |
NO |
|
YES |
YES |
YES |
YES 1 |
YES 1 |
NO |
|
NO |
NO |
YES |
YES 1 |
YES |
NO |
|
NO |
NO |
YES |
YES 1 |
YES |
NO |
|
NO |
NO |
YES |
YES 1 |
YES |
YES 2 |
Exceptions
-
1
SolrTLSandSolrClientTLSare not supported -
2 The
SolrPrometheusExporteris not supported, becausesolr-exporterwas removed in Solr 10
Please note that this represents basic compatibility with the Solr Operator. There may be options and features that require newer versions of Solr. (e.g. S3/GCS Backup Support)
Please test to make sure the features you plan to use are compatible with the version of Solr you choose to run.
Upgrade Instructions
Please follow the instructions below when upgrading the Solr Operator.
However, no matter the upgrade, always read the Upgrade Warnings and Notes for every version between
Upgrading minor versions (vX.A. → vX.B.)
In order to upgrade minor versions (e.g. v0.2.5 → v0.3.0), you must upgrade one minor version at a time (e.g. v0.2.0 → v0.3.0 → v0.4.0).
It is also necessary to upgrade to the latest patch version before upgrading to the next minor version.
Therefore if you are running v0.2.5 and you want to upgrade to v0.3.0, you must first upgrade to v0.2.8 before upgrading to v0.3.0.
There when upgrading minor versions, there may be CRD options that have been deprecated. The Solr Operator will automatically change the existing Solr resources to use the new format. Therefore when upgrading to the next version, which may remove the deprecated option, the existing resources will behave correctly. However, creating new resources with the deprecated options will fail. Be sure to carefully read the Upgrade Warnings and Notes to be sure you are not using deprecated options.
Upgrading patch versions (vX.Y.A → vZ.Y.B)
You should be able to upgrade from a version to any patch version with the same minor and major versions.
It is always encouraged to upgrade to the latest patch version of the minor and major version you are running.
There is no need to upgrade one patch version at a time (e.g. v0.2.5 → v0.2.6 → v0.2.7 → v0.2.8),
instead you can leap to the latest patch version (e.g. v0.2.5 → v0.2.8).
Where the Helm charts are published
The Solr Operator and Solr Helm charts are published as OCI artifacts to the GitHub Container Registry, at oci://ghcr.io/apache/solr-operator/helm.
They can be installed either straight from the registry, or through the Helm repository at https://solr.apache.org/charts, which lists the same charts.
Both are supported.
-
OCI Charts
-
Charts Repo
Installing from the OCI registry needs no helm repo add, and no credentials.
helm install solr-operator oci://ghcr.io/apache/solr-operator/helm/solr-operator --version 0.10.0
helm install example oci://ghcr.io/apache/solr-operator/helm/solr --version 0.10.0
The Helm repository’s entries point at the OCI registry, so it serves exactly the same charts.
It requires Helm v3.8 or later, and keeps existing helm repo add setups and chart dependencies on https://solr.apache.org/charts working.
helm repo add apache-solr https://solr.apache.org/charts
helm repo update
helm install solr-operator apache-solr/solr-operator --version 0.10.0
helm install example apache-solr/solr --version 0.10.0
The signed chart tarballs on the Apache download mirrors remain the official release artifacts. The OCI copies are byte-identical to them, and carry the same signatures, so the charts can be verified with either source. See Verifying the Helm charts.
Verifying the Helm charts
The charts are signed with the Solr release signing keys, and Helm can check that signature when installing or upgrading, with --verify.
Helm reads keys from a keyring in the legacy GnuPG format, ~/.gnupg/pubring.gpg by default.
GnuPG 2.1 and later no longer write that file, so import the Solr keys and export them to a keyring for Helm first:
curl -sL https://dist.apache.org/repos/dist/release/solr/KEYS | gpg --import
gpg --export > ~/solr-keyring.gpg
Then pass --verify and the keyring to helm install or helm upgrade:
-
OCI Charts
-
Charts Repo
helm install solr-operator oci://ghcr.io/apache/solr-operator/helm/solr-operator --version 0.10.0 \
--verify --keyring ~/solr-keyring.gpg
helm install example oci://ghcr.io/apache/solr-operator/helm/solr --version 0.10.0 \
--verify --keyring ~/solr-keyring.gpg
helm install solr-operator apache-solr/solr-operator --version 0.10.0 \
--verify --keyring ~/solr-keyring.gpg
helm install example apache-solr/solr --version 0.10.0 \
--verify --keyring ~/solr-keyring.gpg
Helm refuses to install a chart whose signature does not verify against the keyring.
Installing the Solr Operator vs Solr CRDs
Installing the Solr Operator, especially via the Helm Chart, does not necessarily mean that you are installing the required CRDs for that version of the Solr Operator.
If the Solr CRDs already exist in your Kubernetes cluster, then Helm will not update them even if the CRDs have changed between the Helm chart versions. Instead, you will need to manually install the CRDs whenever upgrading your Solr Operator. You should always upgrade your CRDs before upgrading the Operator
You can do this via the following command, replacing <version> with the version of the Solr Operator you are installing:
# Just replace the Solr CRDs
kubectl replace -f "http://solr.apache.org/operator/downloads/crds/<version>/all.yaml"
# Just replace the Solr CRDs and all CRDs it might depend on (e.g. ZookeeperCluster)
kubectl replace -f "http://solr.apache.org/operator/downloads/crds/<version>/all-with-dependencies.yaml"
It is strongly recommended to use kubectl create or kubectl replace, instead of kubectl apply when creating/updating CRDs.
Upgrading the Zookeeper Operator
When upgrading the Solr Operator, you may need to upgrade the Zookeeper Operator at the same time. If you are using the Solr Helm chart to deploy the Zookeeper operator, then you won’t need to do anything besides installing the CRD’s with dependencies, and upgrade the Solr Operator helm deployment.
# Just replace the Solr CRDs and all CRDs it might depend on (e.g. ZookeeperCluster)
kubectl replace -f "http://solr.apache.org/operator/downloads/crds/v0.10.0/all-with-dependencies.yaml"
helm upgrade solr-operator oci://ghcr.io/apache/solr-operator/helm/solr-operator --version 0.10.0
Note that the Helm chart version does not contain a v prefix, which the downloads version does. The Helm chart version is the only part of the Solr Operator release that does not use the v prefix.
Upgrade Warnings and Notes
v0.10.0
-
Solr 10 support, and Solr image tags must start with the Solr version
The Solr Operator now supports Solr 10. Solr 10 needs a different
solr.xml, different env vars and differentbin/solrcommands than Solr 9, so the operator reads the Solr version fromspec.solrImage.tag. The tag must therefore start with the Solr version, e.g.9.10.0,10.0.0,9.10,8.11,10,9.10.0-slim,10.0.0-my-custom-buildor10.0.0@sha256:<digest>. Tags without a version, such aslatestorslim, and custom tags that don’t start with the Solr version, are not accepted. The operator does not reconcile a SolrCloud with such a tag, or with an unsupported Solr major version; it emits anUnsupportedSolrVersionWarning event instead. Before upgrading the operator, check the image tags of your SolrClouds (including theimage.tagvalue of thesolrHelm chart) and change them where needed. SolrClouds that never set a tag got the operator’s default tag at the time they were created (e.g.8.11), which is accepted. Upgrading the operator restarts the pods of every SolrCloud, including the ones that stay on Solr 8.11 or 9.x. The Solr 10 specific changes are gated on the Solr major version and do not affect them, but other changes in this release do alter the pod template of every SolrCloud: the newsolr.apache.org/storageMinimumSizeannotation, the removal of thepostStartlifecycle hook, and the newSOLR_TOOL_OPTSenvironment variable on the ZooKeeper setup initContainer. This is a managed rolling restart that takes the cluster operation lock and respectsmaxPodsUnavailable, so it behaves like any other rolling restart, but plan for it. Changing a tag, e.g. fromlatestto9.10.0, changes the pod template and therefore restarts all Solr pods, even if the new tag points to the same image. If you pass-Dsolr.allowPaths=…viasolrOpts, use-Dsolr.security.allow.paths=…when moving to Solr 10, which doesn’t read the old camelCase name anymore. Alternatively, set the paths via an environment variable inspec.customSolrKubeOptions.podOptions.envVars:SOLR_ALLOW_PATHSon Solr 9.6+, andSOLR_SECURITY_ALLOW_PATHSon Solr 10. Solr 10 still readsSOLR_ALLOW_PATHS, but logs a deprecation warning for it. Solr 10 can’t open indexes that were created with Solr 8, as Lucene only reads indexes created by the previous major version, so check Solr’s own upgrade notes before upgrading a SolrCloud that started on Solr 8. -
Solr 10: custom
solr.xmlfiles need to be updatedIf you provide your own
solr.xml(viaspec.customSolrKubeOptions.configMapOptions.providedConfigMap), it needs to change together with the Solr version. Solr 10 refuses to start with settings it no longer supports, such ashostContextandgenericCoreNodeNames, while Solr 9 requireshostContext, so no singlesolr.xmlworks for both. Create a new ConfigMap with a Solr 10solr.xml, and changespec.solrImage.tagandprovidedConfigMapin the same update of the SolrCloud. The operator emits anInvalidConfigMapWarning event if the customsolr.xmlof a Solr 10 SolrCloud still containshostContextorgenericCoreNodeNames, but it doesn’t stop the rolling restart. The easiest way is to start from thesolr.xmlthat ships with Solr 10 (server/solr/solr.xml) and re-apply your changes; use${solr.host.advertise:}for thehostsetting. -
Solr 10: custom
log4j2.xmlfiles need to be updatedSolr 10 renamed the
solr.log.dirsystem property tosolr.logs.dir. If you provide your ownlog4j2.xml(viaspec.customSolrKubeOptions.configMapOptions.providedConfigMap), any appender referring to${sys:solr.log.dir}stops working on Solr 10. Log4j cannot resolve the old property, so every appender using it fails to initialise and Solr starts with console logging only, writing no log files at all. Solr still starts and still logs to stdout, which makes this easy to miss. Use${sys:solr.logs.dir}, or${env:SOLR_LOGS_DIR}, which the Solr container image sets on both Solr 9 and Solr 10 and therefore needs no change when upgrading. -
The
SolrPrometheusExporterdoes not support Solr 10solr-exporterwas removed from the Solr distribution in Solr 10, so Solr 10 images don’t contain/opt/solr/prometheus-exporteranymore. ASolrPrometheusExporterwith a Solr 10 image cannot start, and one with an older image cannot read the metrics of Solr 10, which only serves them in Prometheus format. The operator does not create or update the resources of aSolrPrometheusExporterthat uses a Solr 10 image or refers to a Solr 10 SolrCloud; it emits anUnsupportedSolrVersionWarning event and sets its status to not ready instead. An existing exporter Deployment keeps running when its SolrCloud is upgraded to Solr 10, but it no longer reports Solr’s metrics. Delete anySolrPrometheusExporterfor a SolrCloud before upgrading that SolrCloud to Solr 10. Instead, each Solr 10 node exposes its metrics in Prometheus format at/solr/admin/metrics, so Prometheus can scrape the Solr pods directly (see #820). -
The Helm charts are now published as OCI artifacts to
oci://ghcr.io/apache/solr-operator/helm. The Helm repository athttps://solr.apache.org/chartskeeps working, and now lists the same OCI charts, which requires Helmv3.8or later. See Where the Helm charts are published. -
The default Solr version has been upgraded from
8.11to9.11.0. This only affects newly created resources, since the operator writes the resolved Solr version into the resource’s spec on the first reconcile, so an operator upgrade never changes the Solr version of a running cluster. Note that a Solr>=9.10node refuses to start if itsmajor.minorversion is lower than the highest version already running in the same Solr cluster, so a rolling downgrade of Solr requires setting theSOLR_CLOUD_DOWNGRADE_ENABLEDenvVar totrueon the pods. Solr9.11.0carries several CVE fixes, so existing clusters on an older Solr version have to set their image tag themselves to pick them up. -
The CRDs gain the new optional pod fields from the Kubernetes
1.32to1.37API types, settable throughpodOptions. Upgrade the CRDs along with the operator, see Installing the Solr Operator vs Solr CRDs, before using them. No existing field or default changed. -
Operator logs now default to
jsonencoding atinfolevel. Set the new helm valuedevelopmenttotrue(CLI flag--zap-devel) for the previousconsoleencoding atdebuglevel. -
The
cp-solr-xmlinitContainer has been removed. Thesolr.xmlis now mounted read-only into/var/solr/data/solr.xmlfrom the ConfigMap, via asubPathvolumeMount, instead of being copied there before Solr starts.-
This changes the Solr pod spec, so the Solr pods are restarted when the operator is upgraded. SolrClouds using
updateStrategy.method: Manualare the exception, since neither the operator nor Kubernetes deletes their pods: they keep running with the old layout until you delete the pods yourself. -
The
solr.xmlis now only visible inside the Solr container. Sidecars and custom initContainers that mount the data volume no longer get it from there. With persistent orhostPathdata storage they do still see thesolr.xmlthatcp-solr-xmlcopied onto the volume before the upgrade, but that copy is stale and will never be updated again. Mount thesolr.xmlConfigMap into such a container if it needs the current configuration. -
The permission fixing that
cp-solr-xmlalso did forvolumebackupRepositorieshas moved to a newvolume-prepinitContainer, which is only created for SolrClouds that have avolumebackupRepository. All other Solr pods now start without any volume prep initContainer.
-
-
Solr pods now get the
availability_zonesystem property, set from thetopology.kubernetes.io/zonelabel of the node they run on, for Solr’s AffinityPlacementFactory. The label is only copied onto pods by Kubernetesv1.35+, and the property is empty on older versions, which leaves replica placement unchanged. Solr pods also get thespread_domainsystem property, set to the name of their Kubernetes node, which Solr only uses whenspreadAcrossDomainsis enabled in theAffinityPlacementFactoryconfiguration. These are part of the same pod spec change that restarts the Solr pods during this upgrade, so they do not cause an additional restart. -
The
preStophook runningsolr stophas been removed, along with theSOLR_STOP_WAITenvVar that only served it, so pod deletions no longer emit aFailedPreStopHookwarning event. Solr now shuts down on theSIGTERMthat Kubernetes sends, which runs the same graceful shutdown.-
This is part of the same pod spec change that restarts the Solr pods during this upgrade, so it does not cause an additional restart. The old pods emit the
FailedPreStopHookevent one last time as they are replaced. -
A stopped Solr container now exits with code
143instead of0, so its last state showsErrorinstead ofCompleted. -
A custom Solr image whose entrypoint does not pass
SIGTERMon to the Solr process is now killed at the end ofterminationGracePeriodSeconds. Fix the entrypoint, or set your ownpreStophook throughcustomSolrKubeOptions.podOptions.lifecycle.
-
-
Authenticated probes now actually fail when authentication fails
When
solrSecurity.probesRequireAuthis enabled, the generated liveness/readiness/startup probe command used to discard the exit code of the Solr request, so the probes reported success no matter what Solr answered. The probes now exit with the status of the Solr request.If a SolrCloud is running with credentials that the probe endpoints reject, its pods were previously reported as healthy and will now correctly go un-ready and be restarted. Check that the basic auth secret is valid, and that the probe paths are authorized for that user, before upgrading.
v0.8.0
-
The minimum supported Solr version is now 8.11
If you are unable to use a newer version of Solr, please install the
v0.7.1version of the Solr Operator. However, it is strongly suggested to upgrade to newer versions of Solr that are actively supported. See the version compatibility matrix for more information. -
Kubernetes support is now limited to 1.22+.
If you are unable to use a newer version of Kubernetes, please install the
v0.7.1version of the Solr Operator for use with Kubernetes1.21. See the version compatibility matrix for more information. -
The new
SolrCloud.spec.scaling.vacatePodsOnScaleDownoption is enabled by default. This means that any SolrCloud that has itsspec.replicasdecreased will have the replicas migrated off of the soon-to-be-deleted pods by default. Set this value tofalseto retain the previous functionality. More information can be found in the Solr Pod Scale-Down documentation. -
The
POD_HOSTNAMEenvVar in SolrCloud Pods has been deprecated. UsePOD_NAMEinstead. -
Use of the
hostPortsystem property placeholder in custom solr.xml files has been deprecated. Use<int name="hostPort">${solr.port.advertise:80}</int>, the default value used by Solr, instead. -
By default
solrcloudresources will now use/admin/info/systemand/admin/info/healthfor liveness and readiness checks, respectively. Administrators that provide customsecurity.jsonfiles for their clusters should either exempt both of these endpoints from authentication entirely, or configure permissions ensuring the relevant Solr user account can access them without issue.
v0.7.0
-
Kubernetes support is now limited to 1.21+.
If you are unable to use a newer version of Kubernetes, please install the
v0.6.0version of the Solr Operator for use with Kubernetes1.20and below. See the version compatibility matrix for more information. -
The required version of the Zookeeper Operator to use with this version has been upgraded from
v0.2.14tov0.2.15. If you use the Solr Operator helm chart, then by default the new version of the Zookeeper Operator will be installed as well. Refer to the helm chart documentation if you want to manage the Zookeeper Operator installation yourself. Please refer to the Zookeeper Operator release notes before upgrading. Make sure to install the correct version of the Zookeeper Operator CRDs, as shown above. -
PodDisruptionBudgetsare now created by default alongside SolrCloud instances. The maximum number of pods allowed down at any given time is aligned with the Managed Update settings provided in the spec. If this is not provided, the default setting (25%) is used.PodDisruptionBudgetcreation can be disabled for a solrcloud resource, by settingspec.availability.podDisruptionBudget.enabledto false. -
Provided Zookeeper pods use the
IfNotPresentpullPolicy by default. Users that specify this field manually will not see a change. -
The Solr Operator now tries to limit connectivity to pods before they are deleted, for rolling updates or other reasons. Before the pod is killed, and evicted of replicas if ephemeral storage is used, a readinessCondition will be set to
false. The Headless Service does not use readiness, so internode traffic will not be affected, however the ClusterIP (common) service will no longer include these nodes until they have been restarted. This change will improve request success rates during a rolling restart. Refer to the Managed Updates documentation. -
The deprecated SolrCloud CRD field
Spec.solrAddressability.external.ingressTLSTerminationSecrethas been removed, please useSpec.solrAddressability.external.ingressTLSTermination.tlsSecretinstead. In order to have these fields changed automatically, upgrade to thev0.6.0version of the Solr operator before upgrading to thev0.7.0version. However, all new SolrCloud resources will need to respect this field change. -
The deprecated SolrCloud CRD field
Spec.solrAddressability.external.additionalDomainshas been removed, please useSpec.solrAddressability.external.additionalDomainNamesinstead. In order to have these fields changed automatically, upgrade to thev0.6.0version of the Solr operator before upgrading to thev0.7.0version. However, all new SolrCloud resources will need to respect this field change.
v0.6.0
-
The default Solr version for the
SolrCloudandSolrPrometheusExporterresources has been upgraded from8.9to8.11. This will not affect any existing resources, as default versions are hard-written to the resources immediately. Only new resources created after the Solr Operator is upgraded tov0.6.0will be affected. -
The required version of the Zookeeper Operator to use with this version has been upgraded from
v0.2.12tov0.2.14. If you use the Solr Operator helm chart, then by default the new version of the Zookeeper Operator will be installed as well. Refer to the helm chart documentation if you want to manage the Zookeeper Operator installation yourself. Please refer to the Zookeeper Operator release notes before upgrading. Make sure to install the correct version of the Zookeeper Operator CRDS, as shown above. -
The SolrCloud CRD field
Spec.solrAddressability.external.additionalDomainshas been renamed toadditionalDomainNames. In this releaseadditionalDomainsis still accepted, but all values will automatically be added toadditionalDomainNamesand the field will be set tonilby the operator. TheadditionalDomainsoption will be removed in a future version. -
The SolrCloud CRD field
Spec.solrAddressability.external.ingressTLSTerminationSecrethas been moved toSpec.solrAddressability.external.ingressTLSTermination.tlsSecret. In this releaseingressTLSTerminationSecretis still accepted, but all values will automatically be changed toingressTLSTermination.tlsSecretand the original field will be set tonilby the operator. TheingressTLSTerminationSecretoption will be removed in a future version. -
SolrPrometheusExporterresources without any image specifications (SolrPrometheusExporter.Spec.image.) will use the referencedSolrCloudimage, if the reference is byname, notzkConnectionString. If anySolrPrometheusExporter.Spec.image.option is provided, then those values will be defaulted by the Solr Operator and theSolrCloudimage will not be used. When upgrading fromv0.5.*tov0.6.0, only newSolrPrometheusExporterresources will use this new feature. To enable it on existing resources, update the resources and remove theSolrPrometheusExporter.Spec.imagesection. -
CRD options deprecated in
v0.5.0have been removed. This includes fieldSolrCloud.spec.dataStorage.backupRestoreOptions,SolrBackup.spec.persistenceandSolrBackup.status.persistenceStatus. Upgrading tov0.5.will remove these options on existing and new SolrCloud and SolrBackup resources. However, once the Solr CRDs are upgraded tov0.6.0, you will no longer be able to submit resources with the options listed above. Please migrate your systems to use the new options while runningv0.5., before upgrading tov0.6.0.
v0.5.0
-
Due to the deprecation and removal of
networking.k8s.io/v1beta1in Kubernetes v1.22,networking.k8s.io/v1will be used for Ingresses.This means that Kubernetes support is now limited to 1.19+.
If you are unable to use a newer version of Kubernetes, please install the
v0.4.0version of the Solr Operator for use with Kubernetes 1.18 and below. See the version compatibility matrix for more information.This also means that if you specify a custom
ingressClassvia an annotation, you should change to use theSolrCloud.spec.customSolrKubeOptions.ingressOptions.ingressClassNameinstead. The ability to set the class through annotations is now deprecated in Kubernetes and will be removed in future versions. -
The legacy way of specifying a backupRepository has been DEPRECATED. Instead of using
SolrCloud.spec.dataStorage.backupRestoreOptions, useSolrCloud.spec.backupRepositories. TheSolrCloud.spec.dataStorage.backupRestoreOptionsoption will be removed inv0.6.0.Note: Do not take backups while upgrading from the Solr Operator
v0.4.0tov0.5.0. Wait for the SolrClouds to be updated, after the Solr Operator is upgraded, and complete their rolling restarts before continuing to use the Backup functionality. -
The location of Solr backup data as well as the name of the Solr backups have been changed, when using volume repositories. Previously the name of the backup (in solr) was set to the name of the collection. Now the name given to the backup in Solr will be set to
<backup-resource-name>-<collection-name>, without the<or>characters, where thebackup-resource-nameis the name of the SolrBackup resource.The directory in the Read-Write-Many Volume, required for volume repositories, that backups are written to is now
/cloud/<solr-cloud-name>/backupsby default, instead of/cloud/<solr-cloud-name>/backups/<backup-name>. Because the backup name in Solr uses both the SolrBackup resource name and the collection name, there should be no collisions in this directory. However, this can be overridden using theSolrBackup.spec.locationoption, which is appended to/cloud/<solr-cloud-name>. -
The SolrBackup persistence option has been removed as of
v0.5.0. Users should plan to keep their backup data in the shared volume if using a Volume Backup repository. IfSolrBackup.spec.persistenceis provided, it will be removed and written back to Kubernetes.Users using the S3 persistence option should try to use the S3 backup repository instead. This requires Solr 8.10 or higher.
-
Default ports when using TLS are now set to 443 instead of 80. This affects
solrCloud.Spec.SolrAddressability.CommonServicePortandsolrCloud.Spec.SolrAddressability.CommonServicePortfield defaulting. Users already explicitly setting these values will not be affected.
v0.4.0
-
The required version of the Zookeeper Operator to use with this version has been upgraded from
v0.2.9tov0.2.12. If you use the Solr Operator helm chart, then by default the new version of the Zookeeper Operator will be installed as well. Refer to the helm chart documentation if you want to manage the Zookeeper Operator installation yourself. Please refer to the Zookeeper Operator release notes before upgrading. Make sure to install the correct version of the Zookeeper Operator CRDS, as shown above. -
The deprecated Solr Operator Helm chart option
useZkOperatorhas been removed, usezookeeper-operator.useinstead.Note: The old option takes a string
"true"/"false", while the new option takes a booleantrue/false. -
The default Solr version for
SolrCloudandSolrPrometheusExporterresources has been upgraded from7.7.0to8.9. This will not affect any existing resources, as default versions are hard-written to the resources immediately. Only new resources created after the Solr Operator is upgraded tov0.4.0will be affected. -
In previous versions of the Solr Operator, the provided Zookeeper instances could only use Persistent Storage. Now ephemeral storage is enabled, and used by default if Solr is using ephemeral storage. The ZK storage type can be explicitly set via
Spec.zookeeperRef.provided.ephemeralorSpec.zookeeperRef.provided.persistence, however if neither is set, the Solr Operator will default to use the type of storage (persistent or ephemeral) that Solr is using.This means that the default Zookeeper Storage type can change for users using ephemeral storage for Solr. If you require ephemeral Solr storage and persistent Zookeeper Storage, be sure to explicitly set that starting in
v0.4.0.
v0.3.0
-
All deprecated CRD fields and Solr Operator options from
v0.2.*have been removed. -
The
SolrCollectionandSolrCollectionAliashave been removed. Please use the Solr APIs to manage these resources instead. Discussion around the removal can be found in Issue #204. -
The required version of the Zookeeper Operator to use with this version has been upgraded from
v0.2.6tov0.2.9. If you use the Solr Operator helm chart, then by default the new version of the Zookeeper Operator will be installed as well. Refer to the helm chart documentation if you want to manage the Zookeeper Operator installation yourself. Please refer to the Zookeeper Operator release notes before upgrading.
v0.2.7
-
Due to the addition of possible sidecar/initContainers for SolrClouds, the version of CRDs used had to be upgraded to
apiextensions.k8s.io/v1.This means that Kubernetes support is now limited to 1.16+.
If you are unable to use a newer version of Kubernetes, please install the
v0.2.6version of the Solr Operator for use with Kubernetes 1.15 and below. -
The location of backup-restore volume mounts in Solr containers has changed from
/var/solr/solr-backup-restoreto/var/solr/data/backup-restore. This change was made to ensure that there were no issues using the backup API with solr 8.6+, which restricts the locations that backup data can be saved to and read from. This change should be transparent if you are merely using the SolrBackup CRD. All files permissions issues with SolrBackups should now be addressed. -
The default
PodManagementPolicyfor StatefulSets has been changed toParallelfromOrderedReady. This change will not affect existing StatefulSets, asPodManagementPolicycannot be updated. In order to continue usingOrderedReadyon new SolrClouds, please use the following setting:SolrCloud.spec.customSolrKubeOptions.statefulSetOptions.podManagementPolicy -
The
SolrCloudandSolrPrometheusExporterservices' portNames have changed to"solr-client"and"solr-metrics"from"ext-solr-client"and"ext-solr-metrics", respectively. This is due to a bug in Kubernetes whereportNameandtargetPortmust match for services. -
Support for
etcd/zetcddeployments has been removed. The section for a Zookeeper cluster SpecSolrCloud.spec.zookeeperRef.provided.zookeeperhas been DEPRECATED. The same fields (except for the deprecatedpersistentVolumeClaimSpecoption) are now available underSolrCloud.spec.zookeeperRef.provided. -
Data Storage options have been expanded, and moved from their old locations.
-
SolrCloud.spec.dataPvcSpechas been DEPRECATED. Please instead use the following instead:SolrCloud.spec.dataStorage.persistent.pvcTemplate.spec=<spec> -
SolrCloud.spec.backupRestoreVolumehas been DEPRECATED. Please instead use the following instead:SolrCloud.spec.dataStorage.backupRestoreOptions.Volume=<volume-source>
-
v0.2.6
-
The solr-operator argument
--ingressBaseDomainhas been DEPRECATED. In order to set the external baseDomain of your clouds, please begin to useSolrCloud.spec.solrAddressability.external.domainNameinstead. You will also need to setSolrCloud.spec.solrAddressability.external.methodtoIngress. The--ingressBaseDomainargument is backwards compatible, and all existing SolrCloud objects will be auto-updated once your operator is upgraded tov0.2.6. The argument will be removed in a future version (v0.3.0).
v0.2.4
-
The default supported version of the Zookeeper Operator has been upgraded to
v0.2.6. If you are using the provided zookeeper option for your SolrClouds, then you will want to upgrade your zookeeper operator version as well as the version and image of the zookeeper that you are running. You can find examples of the zookeeper operator as well as solrClouds that use provided zookeepers in the examples directory. Please refer to the Zookeeper Operator release notes before upgrading.
v0.2.3
-
If you do not use an ingress with the Solr Operator, the Solr Hostname and Port will change when upgrading to this version. This is to fix an outstanding bug. Because of the headless service port change, you will likely see an outage for inter-node communication until all pods have been restarted.
v0.2.2
-
SolrCloud.spec.solrPodPolicyhas been DEPRECATED in favor of theSolrCloud.spec.customSolrKubeOptions.podOptionsoption. This option is backwards compatible, but will be removed in a future version (v0.3.0). -
SolrPrometheusExporter.spec.solrPodPolicyhas been DEPRECATED in favor of theSolrPrometheusExporter.spec.customKubeOptions.podOptionsoption. This option is backwards compatible, but will be removed in a future version (v0.3.0).
v0.2.1
-
The zkConnectionString used for provided zookeepers changed from using the string provided in the
ZkCluster.Status, which used an IP, to using the service name. This will cause a rolling restart of your solrs using the provided zookeeper option, but there will be no data loss.
v0.2.0
-
Uses
gomodinstead ofdep -
SolrCloud.spec.zookeeperRef.provided.zookeeper.persistentVolumeClaimSpechas been DEPRECATED in favor of theSolrCloud.zookeeperRef.provided.zookeeper.persistenceoption. This option is backwards compatible, but will be removed in a future version (v0.3.0). -
An upgrade to the ZKOperator version
0.2.4is required.