Skip to content
Percona Documentation

For help, click the link below to get free database assistance or contact our experts for personalized support.

Get help from Percona

Valkey 8.1.0 RC1 release notes

Upgrade urgency levels

  • LOW: No need to upgrade unless there are new features you want to use.
  • MODERATE: Program an upgrade of the server, but it’s not urgent.
  • HIGH: There is a critical bug that may affect a subset of users. Upgrade!
  • CRITICAL: There is a critical bug affecting MOST USERS. Upgrade ASAP.
  • SECURITY: There are security fixes in the release.

Valkey 8.1.0 RC1 - Released Tue 11 Feb 2025

Upgrade urgency LOW.

This is the first release candidate of Valkey 8.1, with performance improvements, extended observability and cluster improvements and different bug fixes. It includes a new implementation of the Valkey dictionary which is more memory and cache efficient, better performance for encryption in transit, reduced replication overhead by offloading work to I/O threads, faster failover support in cluster mode, major improvements to the active defrag process to reduce the impact on command processing, different API changes for improved usability and ability to track large requests and replies. Valkey now supports new new check-and-set feature for native STRINGs.

API and Interface changes

  • Introduce cancel argument to bgsave command (#757)
  • Add conditional update support to the SET command using IFEQ argument (#1324)
  • Add more filters to CLIENT LIST (#1401)
  • Add availability_zone to the HELLO response (#1487)

Observability and Monitoring changes

  • Extend LATENCY LATEST to add sum / cnt stats (#1570)
  • Add paused_actions and paused_timeout_milliseconds for INFO CLIENTS (#1519)
  • Add paused_reason to INFO CLIENTS (#1564)
  • Added COMMANDLOG to record slow executions and large requests/replies (#1294)
  • Fix cluster info sent stats for message with light header (#1563)
  • Add latency stats around cluster config file operations (#1534)
  • Add new flag in CLIENT LIST for import-source client (#1398)
  • Show client capabilities in CLIENT LIST / CLIENT INFO (#1698)

Performance/Efficiency Improvements - Core

  • Introduce a new memory efficient hash table to store keys (#1186)
  • Accelerate hash table iterator with prefetching (#1501)
  • Accelerate hash table iterator with value prefetching (#1568)
  • Replace dict with new hashtable: hash datatype (#1502)
  • Replace dict with new hashtable for sets datatype (#1176)
  • Replace dict with new hashtable: sorted set datatype (#1427)
  • Free strings during BGSAVE/BGAOFRW to reduce copy-on-write (#905)
  • Create an empty lua table with specified initial capacity as much as possible (#1092)
  • Move prepareClientToWrite out of loop for HGETALL command (#1119)
  • Improved hashing algorithm for Lua tables (#1168)
  • Replace dict with new hashtable for sets datatype (#1176)
  • Do security attack check only when command not found to reduce the critical path. (#1212)
  • Trim free space from inline command argument strings to avoid excess memory usage (#1213)
  • Increase the max number of io threads to 256. (#1220)
  • Refactor of ActiveDefrag to reduce latencies (#1242)
  • Integrate fast_float to optionally replace strtod (#1260)
  • Improvements for TLS with I/O threads (#1271)
  • Optimize PFCOUNT, PFMERGE command by SIMD acceleration (#1293)
  • Optimize sdscatrepr by batch processing printable characters (#1342)
  • Optimize ZRANK to avoid path comparisons (#1389)
  • Move clientCron onto a separate timer (#1387)
  • Client struct: lazy init components and optimize struct layout (#1405)
  • Offload reading the replication stream to IO threads (#1449)
  • Skip CRC checksumming during diskless full sync with TLS enabled. (#1479)

New/Modified configurations

  • Deprecate io-threads-do-reads, which has no effect since io threads will now always do reads. (#1138)
  • Introduce import-mode config to avoid expiration and eviction during data syncing (#1185)
  • Introduce new rdb-version-check config which allows for relaxed RDB version verification (#1604)
  • Deprecate dynamic-hz, since server cron jobs are handled dynamically by default (#1387)
  • Introduce log-format and log-timestamp-format to control the log format (#1022)
  • Introducing active-defrag-cycle-us for more fine-grinned control of memory defragmentation run time (#1242)
  • Introduce new configurations to control the new COMMANDLOG reporting thresholds (#1294)

Build and Packaging changes

  • Introduce CMake build system for valkey (#1196)
  • RDMA builtin support (#1209)
  • Fix Valkey binary build workflow, version support changes. (#1429)
  • Remove Valkey specific changes in jemalloc source code (#1266)

Module Improvements

  • Add API UpdateRuntimeArgs for updating the module arguments during runtime (#1041)
  • Add support for MustObeyClient Module API (#1582)
  • Adds support for scripting engines as Valkey modules (#1277, #1497)

Cluster improvements

  • Do election in order based on failed primary rank to avoid voting conflicts (#1018)
  • Make replica CLUSTER RESET flush async based on lazyfree-lazy-user-flush (#1190)
  • Trigger the election as soon as possible when doing a forced manual failover (#1067)
  • Make manual failover reset the on-going election to promote failover (#1274)
  • Brocast a PONG to all node in cluster when role changed (#1295)
  • Manual failover vote is not limited by two times the node timeout (#1305)
  • Automatic failover vote is not limited by two times the node timeout (#1356)

Behavior Changes

  • Streams use an additional 8 bytes to track their internal size (#688)
  • Take hz into account in activerehashing to avoid CPU spikes (#977)
  • Incr expired_keys if the expiration time is already expired (#1517)
  • Fix replica not able to initiate election in time when epoch fails (#1009)
  • Make FUNCTION RESTORE FLUSH flush async based on lazyfree-lazy-user-flush (#1254)
  • Allow MEMORY MALLOC-STATS and MEMORY PURGE during loading phase (#1317)
  • Use DEEPBIND flag when loading external modules in order to avoid symbol conflicts (#1703)

Logging and Tooling Improvements

  • Remove the restriction that cli –cluster create requires at least 3 primary nodes (#1075)
  • Add short client info log to CLUSTER MEET / FORGET / RESET commands (#1249)
  • Support for reading from replicas in valkey-benchmark (#1392)
  • valkey-cli will now re-select previously selected database after reconnect (#1694)
  • valkey-cli will now auto-exit from subscribed mode when there are no more active subscriptions (#1432)

Bug Fixes

  • Mark the node as FAIL when the node is marked as NOADDR and broadcast the FAIL (#1191)
  • [Bug Fix] Optimize RDB Load Performance and Fix Cluster Mode Resizing (#1199)
  • Log as primary role (M) instead of child process (C) during startup (#1282)
  • Fix empty primary may have dirty slots data due to bad migration (#1285)
  • RDMA: Fix dead loop when transfer large data (20KB) (#1386)