Timeplus Enterprise 3.2

Key Highlights

Key highlights of the Timeplus 3.2 release include:

1. Major performance improvement in the data replication network layer — up to 30x faster in some scenarios — powered by request pooling, recyclable network buffers, sharded request/response channels, scatter/gather writes, and IPv6 support.
2. Major performance improvement (up to 40x) for Kafka consume / parsing for Protobuf, CSV, and similar formats via smart batching and a new parallel parsing strategy for Kafka source.
3. Major enhancements to Python UDFs and external Python table functions now enable secure, direct communication with the local timeplusd instance via automatically provisioned an ephemeral user and token.
4. Broad stability and quality hardening across mutable streams, checkpoints, materialized views, streaming joins, memory accounting, and replicated log recovery.
5. Improved Okta SSO integration with a smoother login flow and support for mapping Okta users to read-only or admin roles.

Supported OS

Deployment Type	OS
Linux bare metal	x64 or ARM chips: Ubuntu 20.04+, RHEL 8+, Fedora 35+, Amazon Linux 2023
Mac bare metal	Intel or Apple chips: macOS 14, macOS 15
Kubernetes	Kubernetes 1.25+, with Helm 3.12+

Releases

We recommend using stable releases for production deployment. Engineering builds are available for testing and evaluation purposes.

3.2.3

Released on 04-12-2026. Installation options:

• For Linux or Mac users: curl https://install.timeplus.com/3.0 | sh Downloads
• For Docker users (not recommended for production): docker run -p 8000:8000 docker.timeplus.com/timeplus/timeplus-enterprise:3.2.3
• For Kubernetes users: helm install timeplus/timeplus-enterprise --version 11.0.7

Component versions:

• timeplusd 3.2.3
• timeplus_appserver 3.2.1
• timeplus_connector 3.2.0
• timeplus cli 3.0.0
• timeplus byoc 1.0.0

Changelog

This release consolidates all timeplusd changes from 3.1.2 through 3.2.3.

Features and Enhancements

• Add global tiered storage policy (#11811)
• Add feature flag to disable the workload rebalancer (#11815)
• Enhance parallel Kafka source (#11742)
• Add Python external stream init and deinit hooks (#11756)
• Add server config gates for JavaScript and Python UDFs (#11752)
• Propagate local API user to Python UDF (#11806) and introduce local API user support (#11804)
• Support non-const arguments for rand distribution functions (#11715)
• Add time-weighted aggregate combinator support and coverage (#11583)
• Support map type generation in random stream (#11586)
• Add keep_range_join_max_buckets setting to cap range join bucket count (#11775)
• Improve discard log messages to include range_time_column name (#11854)
• Add IPv6 support (#11653) and IPv6 enforce setting with misc fixes (#11670)
• Enable client batch send by default (#11658)
• Upgrade curl to 8.18 and enable Pulsar on macOS (#11693)
• Upgrade Pulsar client to v4.0.1 (#11611)
• Backfill right-side hash table in streaming enrichment joins (#11702)
• Tolerate UNKNOWN_STREAM in DROP CASCADE / IF EXISTS (#11820)
• Derive subject name according to schema subject strategy (#11705)
• Better error handling in stream metadata updates (#11542)
• Move external stream validation from constructors to dedicated methods (#11636)
• Validate task creation (#11664)

Performance

• Batch process Protobuf messages for Kafka (#11572)
• Avoid Field temporaries in C++ to Python column conversion (#11836)
• Set fill_cache=false on KV full-scan read paths (#11831)
• Add Python GIL wait instrumentation (#11808)
• Improve Python UDF data conversion and Python consume sink performance (#11788)
• Commit historical data inline in the poll thread (#11749)
• Merge small JSON blocks before commit (#11765)
• Add request pooling (#11643)
• Introduce recyclable network buffers (#11626)
• Scatter/gather writes for client (#11629)
• Sharding inflight requests map (#11628) and sharding response channel (#11655)
• Switch blocking queue to dequeue with queue reuse (#11622)
• Optimize eloop wakeup (#11617) and net perf baseline (#11616)
• Fine tune network buffer size (#11641)
• Refactor string-concatenation INSERTs to block-based inserts (#11573)
• Lock guard in consumer (#11800)

Bug Fixes

• Disable checkpoints at runtime for random-source materialized views (#11827)
• Fix join changelog mode for nested aggregation (#11833)
• Fix cgroup memory accounting (#11817)
• Prevent early meta-log compaction (#11826)
• Fix zero replication client (#11813)
• Fix async commit batch and client lifecycle (#11810)
• Fix memory tracker regression causing spurious MEMORY_LIMIT_EXCEEDED (#11783)
• Fix TimeWheel races and shutdown ordering (#11718)
• Fix heap-use-after-free in RocksDBColumnFamilyHandler::destroy() (#11758)
• Clear spilled hybrid update state after emit batches (#11741)
• Fix replicated log startup epoch recovery (#11728)
• Handle Python source cancel during MV failover (#11727)
• Fix context expired MV table subquery (#11738)
• Fix streaming CTE/subquery with aggregation returning empty results (#11695)
• Update MV schema after ALTER STREAM MODIFY QUERY (#11681)
• Keep session watermark monotonic (#11649)
• Support more streaming resize combinations (#11671)
• Fix synthesizeQuorumReplicationStatus (#11668)
• SQL analyzer now returns 400 for invalid SQL (#11708)
• Fix watchdog restart semantics (#11706)
• Preserve OR semantics with empty IN on mutable stream (#11596)
• Handle malformed Confluent Protobuf messages gracefully (#11587)
• Fix avg time-weighted overflow (#11591)
• Fix fatal crash from stale async checkpoint epoch mismatch (#11601)
• Fix secondary index cleanup on delete for mutable stream (#11593)
• Fix server wakeup during shutdown (#11633)
• Fix meta log lagging (#11682)
• Improve FetchHint unset file_position logging (#11772)
• Backport replxx updates (#11818)
• Port Proton OSS #1124: merge small JSON block before commit (#11765)
• Add self-join regression coverage fix (#11678)