...
Improvements and Bug Fixes
Live SPLITSHARD can lose updates due to cluster state change between checking if the current shard is active and later checking if there are any sub-shard leaders to forward the update to
- Fix for SPLITSHARD (async) with failures in underlying sub-operations can result in data loss
- Allow dynamic resizing of SolrCache-s
If a partial update (aka atomic update) is attempted on a document that has child docs, then ensure the schema supports it (_root_ stored/docValues) by throwing an exception
Allow optional redaction of data saved by 'bin/solr autoscaling -save'
Optimized large managed schema modifications (internal O(n^2) problem)
Max idle time support for SolrCache implementations
Add Prometheus Exporter GC and Heap optionsAdd debug, trace logging to RuleBasedAuthorizationPlugin
SSL: Adding Enabling/Disabling client's hostname verification configManaged schema operations should do a core reload in Solr standalone mode
Introducing SolrClient.ping(collection) in SolrJ
Fix for Backup/Restore command supporting picking one of a few repositories by repository parameterFix for CDCR bootstrap not replicating index to the replicas of target cluster
Fixed a race condition when initializing metrics for new security plugins on security.json change
Fixed JWTAuthPlugin to update metrics prior to continuing w/other filters or returning error (hossman)
Fix for SPLITSHARD (async) with failures in underlying sub-operations can result in data loss
Fixed distributed grouping when multiple 'fl' params are specified
JMX MBeans are not exposed because of race condition between creating platform mbean server and registering mbeans
Fix for class-cast issues during atomic-update 'removeregex' operations
Fix for multi-node race condition to create/remove nodeLost markers
Live shard split (where updates actively continue during the split) can lose updates due to cluster state happening to change between checking if the current shard is active and later checking if there are any sub-shard leaders to forward the update to
- Fix for too many cascading calls to remote servers, which can bring down nodesFix for HttpSolrCall producing incorrect extra AuditEvent on AuthorizationResponse.PROMPT
Fix for MOVEREPLICA ignoring replica type and always adding 'nrt' replicas
Fix: DistributedZkUpdateProcessor should propagate URP.finish() lifecycle (regression since 8.1)
...