Introduction

The SolrCloud Search Service Adapter connects the Intershop Commerce Management (ICM) with a Solr server that runs in cloud mode. The search index configuration and maintenance in the ICM back office uses the Schema- and ConfigSet-REST-APIs provided by the newer Solr server version. Thus, the Solr Server is separate and independent of the ICM application server cluster.

References

• Overview - Search

• Concept - Search Services

• Cookbook - Deployment of Solr Search Adapters

• Guide - Deployment Solr Cloud Server

• Guide - Solr Replication

• Guide - Migration Solr4 to SolrCloud Adapter

• Public Release Note - SolrCloud Search Service Adapter 2

Version Information and Dependencies

There is a backward compatibility issue with SolrJ (SOLR-15145). Solr 8.6.3/Solr 8.7+ requires additional configuration: (ISSEA-167 - Can't Create a Configset - Solr Cloud 8.6.3 8.7).

The listed Solr server versions were tested to work with the listed SolrCloud Adapter version. In general, minor and patch version upgrades of the Solr server are usually compatible with the adapter.

The following table lists the cartridges contained in the SolrCloud Adapter:

Deployment

Please see Cookbook - Deployment of Solr Search Adapters on how to set up your project.

The SolrCloud Search Service Adapter requires a Solr server that runs in cloud mode. Please see Guide - Deployment Solr Cloud Server on how to set up a Solr server for development or production. This guide also provides information on how to configure ICM to use a Solr server.

Solr server versions higher than 8.6.2 require setting up authentication or enabling upload of config sets from the ICM application server.

solr.in.sh

SOLR_OPTS="$SOLR_OPTS -Dsolr.disableConfigSetsCreateAuthChecks=true"

Data Replication

The data replication of indexes has changed with this adapter implementation. The standard ICM file replication is not used anymore to replicate the index data. Instead, the adapter comes with its own replication decorator that uses the backup/restore collection API functionality of Solr. Please see Guide - Solr Replication for the description and configuration of the index replication with the SolrCloud server.

Solr Server version 8.6.3 or higher in conjunction with data replication requires setting the backup location as an allowed path for the Solr server cluster. You can specify the allowed paths by setting the system property via environment in the startup script:

solr.in.sh

SOLR_OPTS="$SOLR_OPTS -Dsolr.allowPaths=/mnt/solr_backup"

Changelog

Release 3.7.1

Notes

This release is built with ICM 7.10.41.0 and contains the same bug fixes as release 3.6.3.

Bugfixes

Release 3.6.3

Bugfixes

Release 3.6.2

Bugfixes

Release 3.6.1

Bugfixes

Release 3.6.0

Features

Bugfixes

Release 3.5.2

Bugfixes

Release 3.5.0

Notes

When using custom code that extends or overrides the BackupAndRestoreSolrCloudIndexesDecorator you need to be aware of the changed signature of the switchAliases method. The method switchAliases(HashMap<Domain, Set<String>> processedIndexesByDomain) is deprecated and the new method switchAliases(Map<Domain, Set<String>> processedIndexesByDomain) is used now (87520).

A new property solr.RebuildInconsistentIndexes controls the behavior with inconsistent indexes during the replication of indexes. If the property is set to true, inconsistent indexes are attempted to be rebuilt during the replication. The default is false, so the replication will fail if inconsistent indexes are detected as it was the default behavior before (86951).

Features

Bugfixes

Release 3.4.1

Bugfixes

Release 3.4.0

Notes

Fix 82573 required changing the signature of method public void addHit in class om.intershop.adapter.search_solr.internal.objectsuggest.SuggestCollector from public void addHit(SolrInputDocument inputDoc) to public void addHit(SolrInputDocument inputDoc, Object docFlagValue). Custom implementations that extend this class need to pass the onlineFlag document value as additional parameter to correctly count online hits only.

Features

Bugfixes

Release 3.2.5

Notes

Synonyms are uploaded as full synonyms via _schema_analysis_synonyms_default.json if and only if the Solr server has enabled authentication.

This is checked by the Solr Adapter. An Upload request is used instead of multiple requests if the needed single operations exceed 15 deletes (81264).

Bugfixes

Release 3.2.4

Bugfixes

Release 3.2.3

Bugfixes

Release 3.2.2

Bugfixes

Release 3.2.1

Bugfixes

Release 3.2.0

Features

Configuration of Solr Cloud Server Connection

The Solr server connection configuration can use an alternative property to connect to the server without using the Zookeeper host list. This connection property can contain a single common service endpoint (e.g., a Solr Server Kubernetes setup with Solr Operator) or a list of Solr server nodes.

ac_solr_cloud.properties

# A Solr server URL or a list of Solr server URLs running in SolrCloud mode (separated by ',' or ';')
# Default: none - if not present, solr.zooKeeperHostList is used
#solr.cloudSolrServerURLs=http://localhost:8983/solr

BackupAndRestoreDecorator is bound via Guice (74700)

The deployment of the Solr cloud adapter does not overwrite the staging.properties to add the decorator. The decorator is bound to the processor via Guice framework now.

Updated SolrJ Client Library

We updated the used client library to current version 8.11.1. Older versions will have conflicts with the dependencies of ICM vs SolrJ dependencies.

Bugfixes

Performance of Result Grouping Queries (74431)

The default behavior of counting filter entries for product variations has changed. Due to the heavy performance impact of the group.facet parameter with a higher number of products and higher number of facets, the parameter is not used anymore to calculate the counts. This will affect search index configurations indexing product variations with GroupMasterUUID. In addition, group.facet has also limitations with calculating counts for numeric data types.

However, to retain the previous behavior, it is recommended to add the parameter to the Solr query with code that extends the default SolrIndexExtension implementation via Guice.

Release 3.1.0

This release is based upon ICM 7.10.38 LTS and is not compatible anymore with 7.10.32 LTS releases.

Features

The used query parser changed from Dismax to EDismax. Queries can directly use extended syntax such as wildcards and boolean operators. Fielded search and multi-word synonyms are supported now.

Full index build (re-build) now uses a separate collection to fully build the index beside the currently used online index. Aliasing is used to switch over atomically to the newly built collection. The collection naming scheme has changed to include the time suffix: <cluster-index-prefix><domain-name>-<index-id><yyyyMMddHHmmss>. The previous collection is kept and is deleted after a new successful index re-build. This allows to switch back the alias to the previous collection (manually). Please note as there are now two collections on the Solr server for one index on the ICM side, the resource usage will be higher on the Solr server side.

Bugfixes

Release 3.1 also contains bug fixes from 3.0.x releases up to 3.0.5.

Release 3.0.7

Bugfixes

Release 3.0.6

Bugfixes

Release 3.0.5

To fix issues while closing SolrClient objects in a cache implemented with the common ICM cache framework (71243 AlreadyClosedException), the component SolrClientCache is no longer used and is deprecated.

If customized code is used, SolrClientCache must be replaced with SolrServerProvider instead.

Bugfixes

Release 3.0.4

Bugfixes

Release 3.0.3

 Bugfixes

Release 3.0.1

Bugfixes

Release 3.0.0

This release is dedicated to ICM releases 7.10.31 and higher (Tomcat 9 Release). The adapter release contains the required dependency changes for Tomcat 9. There are no functional or API changes in the adapter compared to adapter release 2.3.4.