We’re happy to announce today the release of Siren Federate version 26, which is compatible with Elasticsearch version 7.15.1. This release includes the following enhancements and fixes:
Enhancements:
- Parts of our tuple-at-a-time query processing pipeline are replaced with a vectorized query processing pipeline. This will improve the overall performance of the system by increasing the worker thread throughput and by reducing in-memory I/O load by using a contiguous memory chunk. Our objective is to complete this work and switch to a fully-vectorized query processing pipeline in Siren Federate version 27. A benchmark report will be published at release time.
- Upgraded to Elasticsearch version 7.15.1.
Breaking changes:
- Removed a deprecated property,
siren.io.netty.maxDirectMemory
, in favor of thesiren.memory.root.limit
setting. - If Search Guard is the configured security system, the node-level setting
searchguard.allow_custom_headers: "siren.*"
must be added to each node of the cluster. - Changed the format of the JSON response for Cluster APIs Nodes Statistics.
Bug fixes:
- Fixed a join cache issue with Elastic Stack Security’s templated queries when they include placeholders such as
{{ _user: user.name }}
. - Fixed a
ConcurrentModificationException
in the select phase following a recent change in Elasticsearch 7.15 involving theFieldUsageTrackingDirectoryReader
. - Adjusted the validation query in the datasource validation endpoint with the latest Avatica JDBC connector.
- Adjusted the autoCommit setting to work with the latest Avatica JDBC connector.
- Fixed a rare issue where a cluster is unable to initialize a Federate job due to a closed
RootBufferManager
on a node. - Fixed a broken link to the
_sample
method in the connector APIs documentation.
Download it here.