What redis/redis shipped
Written by FoxPlug from the repository's own releases; not affiliated with Redis. Paste your own repository in the box below and FoxPlug writes the same weekly update for your product. This page is an automatic summary of the releases, pull requests and commits in github.com/redis/redis; Redis did not write it and does not use or endorse FoxPlug. Every line links to the change it describes.
Get a weekly update like this for your product, free
Week of November 2, 2026
What shipped
- Slot migration import now fails when the destination rejects a command, instead of logging the error and taking over the slots anyway. Pull request #16176
- Fixed stale dictEntryLink reuse after keyspace lookups that could update the wrong key or silently lose a write. Pull request #16121
- Fixed AOF rewrite duplicating commands when an AOF write fails, since buffered commands are no longer re-applied into the new incremental file. Pull request #16175
- Raised a replica's output buffer soft limit to repl-backlog-size, matching the hard limit and what redis.conf describes. Pull request #16098
- Fixed MSETEX propagation with repeated EX or PX options, which had left earlier relative-expiration pairs alongside PXAT in the command sent to AOF and replicas. Pull request #16096
- Client memory now counts watched key and Pub/Sub channel and pattern name objects, so large names show up in tot-mem and against maxmemory-clients. Pull request #16069
- Added per-command CPU time to SLOWLOG, an upper bound on thread CPU time for each logged command. Pull request #15907
- DELEX now honors lazyfree-lazy-user-del instead of reading the server-del flag in both places. Pull request #16120
- SORT and SORT_RO no longer convert listpack-encoded sorted sets to btree, so the source key encoding is left unchanged. Pull request #15622
- CONFIG GET and CONFIG REWRITE keep the n, o and c keyspace event flags when A is set, so they survive a restart. Pull request #16142
Why it matters
This week's changes mostly harden correctness in places users would not expect to break: slot migration imports, AOF rewrite after write errors, client memory accounting and key encoding changes made by read-only commands. Several fixes address silent data problems, such as a reused dict entry link updating the wrong key and duplicate key metadata classes accepted on load. Slowlog now reports per-command CPU time, which gives a way to tell CPU starvation apart from slow commands.
Changelog entry
- Fail slot migration import when the destination rejects a command. Pull request #16176
- Fix stale dictEntryLink reuse after keyspace lookups that could update the wrong key. Pull request #16121
- Fix AOF rewrite duplicating commands after an AOF write error. Pull request #16175
- Raise replica output buffer soft limit to repl-backlog-size. Pull request #16098
- Fix MSETEX propagation with repeated EX or PX options. Pull request #16096
- Account for watched key and Pub/Sub name memory in client memory. Pull request #16069
- Add per-command CPU time to SLOWLOG. Pull request #15907
- Make DELEX honor lazyfree-lazy-user-del. Pull request #16120
- Avoid changing sorted set encoding in SORT and SORT_RO. Pull request #15622
- Keep n, o and c keyspace event flags when A is set. Pull request #16142
- Reject duplicate key metadata classes on load. Pull request #16146
- Avoid empty --keystats reports in redis-cli when no keys are found. Pull request #15823
Weekly fixes: slot migration import now fails on rejected commands, AOF rewrite no longer duplicates commands after a write error, and stale dict entry links can no longer update the wrong key. SLOWLOG also gains per-command CPU time.
This week we fixed several correctness issues. Slot migration import now fails when the destination rejects a command rather than taking over the slots anyway. AOF rewrite no longer duplicates commands after an AOF write error. A stale dict entry link reused across keyspace lookups could update the wrong key or silently lose a write. Replica output buffer soft limits now match repl-backlog-size. MSETEX propagation with repeated EX or PX options is fixed. Client memory now counts watched key and Pub/Sub name objects. SLOWLOG reports per-command CPU time, and DELEX honors lazyfree-lazy-user-del.
Week of October 26, 2026
What shipped
- 8.10.3 shipped with a security fix for a heap buffer overflow when propagating a PUBLISH or SPUBLISH message larger than 4GB over the cluster bus in cluster mode. Release
- 8.8.4 shipped with the same cluster bus PUBLISH and SPUBLISH heap buffer overflow security fix. Release
- 8.6.8 shipped with the same cluster bus PUBLISH and SPUBLISH heap buffer overflow security fix. Release
- 8.4.8 shipped with the same cluster bus PUBLISH and SPUBLISH heap buffer overflow security fix. Release
- 8.2.11 shipped with the same cluster bus PUBLISH and SPUBLISH heap buffer overflow security fix. Release
- 7.4.12 shipped with the same cluster bus PUBLISH and SPUBLISH heap buffer overflow security fix. Release
- 7.2.17 shipped with the same cluster bus PUBLISH and SPUBLISH heap buffer overflow security fix. Release
- 8.12 RC2 was released as a feature-complete pre-release, not suitable for production, updating the Search module with RedisSearch-related content intended for RC1. Release
- A pull request adds per-member expiration to sets, so members of a set can expire individually. Pull request #15938
- A pull request makes redis-cli, redis-benchmark and Sentinel verify the TLS server identity by default, a breaking change for TLS users of these tools. Pull request #16033
Why it matters
Seven maintenance releases across the 8.10, 8.8, 8.6, 8.4, 8.2, 7.4 and 7.2 lines carry a security fix for a cluster bus heap overflow tied to oversized PUBLISH and SPUBLISH propagation, so operators on those lines should plan upgrades. Alongside that, 8.12 RC2 moves the next feature release closer, and open pull requests propose new capabilities such as per-member set expiration.
Changelog entry
- 8.10.3: security fix for heap buffer overflow when propagating PUBLISH/SPUBLISH larger than 4GB over the cluster bus in cluster mode. Release
- 8.8.4: security fix for heap buffer overflow when propagating PUBLISH/SPUBLISH larger than 4GB over the cluster bus in cluster mode. Release
- 8.6.8: security fix for heap buffer overflow when propagating PUBLISH/SPUBLISH larger than 4GB over the cluster bus in cluster mode. Release
- 8.4.8: security fix for heap buffer overflow when propagating PUBLISH/SPUBLISH larger than 4GB over the cluster bus in cluster mode. Release
- 8.2.11: security fix for heap buffer overflow when propagating PUBLISH/SPUBLISH larger than 4GB over the cluster bus in cluster mode. Release
- 7.4.12: security fix for heap buffer overflow when propagating PUBLISH/SPUBLISH larger than 4GB over the cluster bus in cluster mode. Release
- 7.2.17: security fix for heap buffer overflow when propagating PUBLISH/SPUBLISH larger than 4GB over the cluster bus in cluster mode. Release
- 8.12 RC2 released as a feature-complete pre-release, not for production; updates the Search module with RedisSearch-related content intended for RC1. Release
- Fix out-of-bounds read in cluster bus PING/PONG/MEET packet parsing by validating length fields against received bytes. Pull request #16104
- Add per-member expiration to sets. Pull request #15938
- Add BITRANGEINTARRAY for native bitmap enumeration of set or clear bit offsets in a range. Pull request #15986
- Verify the TLS server identity by default in redis-cli, redis-benchmark and Sentinel. Pull request #16033
Seven maintenance releases ship a security fix for a heap overflow when propagating PUBLISH/SPUBLISH messages larger than 4GB over the cluster bus in cluster mode. 8.12 RC2 is also out as a feature-complete pre-release, not for production.
This week brought seven maintenance releases: 8.10.3, 8.8.4, 8.6.8, 8.4.8, 8.2.11, 7.4.12 and 7.2.17. Each carries a security fix for a heap buffer overflow when propagating a PUBLISH or SPUBLISH message larger than 4GB over the cluster bus in cluster mode. If you run cluster mode with pub/sub, review the releases for your line. Also out: 8.12 RC2, a feature-complete pre-release that is not suitable for production, which updates the Search module with the RedisSearch-related content intended for RC1.
Week of October 19, 2026
What shipped
- Large sorted sets now use a packed B+ tree encoding instead of skiplist plus dict, reducing memory for sets that exceed the listpack thresholds. Pull request #15635
- Fixed a remote use-after-free in cluster bus FORGOTTEN_NODE handling that could lead to remote code execution. Pull request #15685
- Fixed Lua shebang and function metadata parsing so malformed input returns an error instead of injecting into the client reply stream. Pull request #16032
- CLUSTER RESET on a replica now propagates the data flush to the AOF and to sub-replicas. Pull request #16038
- MSETEX now has a key-extraction callback, so acknowledged writes are no longer lost during atomic slot migration. Pull request #15937
- Fixed XADD and XTRIM MAXLEN with ACKED stopping early at a node that still has referenced entries. Pull request #16006
- Set TCP_NODELAY and keepalive on the master connection after PSYNC to avoid Nagle delays on REPLCONF ACK. Pull request #16020
- Fixed a Lua denial of service where pattern matching with many combinations in string.find could run indefinitely. Pull request #16017
- Added the io-threads-client-types config to select whether normal, replica or primary connections are handled by IO threads. Pull request #15913
- Client unshared reply memory accounting is now updated incrementally in cron instead of scanning every pending reply buffer. Pull request #15761
Why it matters
This week brings a new memory layout for large sorted sets, which lowers memory use for the biggest sorted set workloads. It also closes several security-relevant bugs, including a cluster bus use-after-free with remote code execution impact, Lua reply injection and heap address exposure, and a Lua denial of service. Replication and cluster paths also get fixes for lost MSETEX writes during slot migration, replica reset persistence, and post-PSYNC socket setup.
Changelog entry
- Large sorted sets now use a packed B+ tree encoding to reduce memory. Pull request #15635
- Fix remote use-after-free in cluster bus FORGOTTEN_NODE handling. Pull request #15685
- Fix Lua shebang and function metadata parsing to return errors instead of injecting into the reply stream. Pull request #16032
- Prevent Lua tostring from exposing process memory addresses. Pull request #16031
- Fix Lua denial of service from long-running string.find pattern matching. Pull request #16017
- CLUSTER RESET on a replica now propagates the flush to the AOF and sub-replicas. Pull request #16038
- Fix MSETEX writes lost during atomic slot migration by adding a key-extraction callback. Pull request #15937
- Fix XADD and XTRIM MAXLEN with ACKED stopping early at a node with referenced entries. Pull request #16006
- Set TCP_NODELAY and keepalive on the master connection after PSYNC. Pull request #16020
- Add io-threads-client-types config to select which connections use IO threads. Pull request #15913
- Make client unshared reply memory accounting incremental in cron. Pull request #15761
- Format ZSCAN skiplist scores with d2string for consistency with ZSCORE. Pull request #15821
This week: large sorted sets get a packed B+ tree encoding to cut memory, plus fixes for a cluster bus use-after-free and Lua reply injection, MSETEX writes during slot migration, and post-PSYNC socket setup.
This week in Redis: large sorted sets now use a packed B+ tree encoding instead of skiplist plus dict, reducing memory for sets beyond the listpack thresholds. Several security fixes landed, including a cluster bus use-after-free in FORGOTTEN_NODE handling with remote code execution impact, Lua shebang and function metadata reply injection, Lua tostring exposing heap addresses, and a Lua string.find denial of service. Cluster and replication paths were also fixed: CLUSTER RESET on a replica now flushes the AOF and sub-replicas, MSETEX writes are no longer lost during atomic slot migration, and TCP_NODELAY and keepalive are set after PSYNC.
Week of October 12, 2026
What shipped
- Redis bitmaps gained a native Roaring bitmap representation backed by vendored CRoaring, which stores set positions more compactly than the existing one-byte-per-position string form. Pull request #15331
- A new INFO Streams section reports per-database base-2 logarithmic histograms of stream and consumer-group properties, in the same shape as INFO keysizes. Pull request #15682
- CRC64 checksumming for RDB load was accelerated with a runtime-dispatched PCLMULQDQ path, about 6x faster for buffers of 64 bytes or more, keeping the existing table implementation as fallback. Pull request #15732
- LZF decompression gained an LZ4-style decoder used for RDB loading, replacing the computed-jump switches and variable-length byte loops with a fast chunk-copying loop. Pull request #15765
- Grisu digit generation in the bundled fpconv dependency no longer performs a 64-bit integer division for every fractional double formatted into a RESP reply. Pull request #15787
- BITOP looks up each source key only once again, fixing the performance regression from the native compressed bitmaps work that had doubled lookups per call. Pull request #15970
- An approximate MINID trim from XTRIM or XADD with trimming options is now rewritten correctly before propagation when tombstones are involved, so replicas and the AOF trim deterministically. Pull request #15936
- Disconnecting after blocked module authentication was processed for unblocking but before AUTH retried now runs the module private-data cleanup callback instead of leaking its blocked-client count. Pull request #15818
- Redis-cli --cluster now accepts the standard bracketed IPv6 form such as [::1]:7000, which previously failed name resolution before connecting. Pull request #15890
- Crash report writing is now bounded so a stuck report no longer prevents the process from exiting. Pull request #15891
Why it matters
This week touches both the format side and the data structures people rely on: Roaring bitmaps change how bitmaps are stored, checksumming and decompression work speeds up RDB load, and a new INFO Streams section exposes consumer-group histograms. Several fixes restore behavior that regressed earlier, including BITOP key lookups and stream trim propagation. Module authors also get corrected cleanup on disconnect and macOS module builds that link against the selected SDK.
Changelog entry
- Add native Roaring bitmap representation for bitmaps, backed by vendored CRoaring. Pull request #15331
- Add INFO Streams section with per-database stream and consumer-group histograms. Pull request #15682
- Accelerate CRC64 checksumming for RDB load with a runtime-dispatched PCLMULQDQ path. Pull request #15732
- Optimize LZF decompression for RDB load with a new LZ4-style decoder. Pull request #15765
- Remove integer divisions from Grisu digit generation in fpconv. Pull request #15787
- Fix BITOP performance regression by looking up each source key only once. Pull request #15970
- Fix stream approximate MINID trim propagation with DELREF and ACKED. Pull request #15936
- Fix blocked module auth cleanup on client disconnect. Pull request #15818
- Fix redis-cli --cluster parsing of bracketed IPv6 addresses. Pull request #15890
- Bound a stuck crash report so the process still exits. Pull request #15891
- Reject DEBUG RELOAD from loading a stale background snapshot. Pull request #15949
- Fix CRC64 test build on ARM when PCLMUL is unavailable. Pull request #15974
This week: native Roaring bitmaps for Redis bitmaps, an INFO Streams section with consumer-group histograms, faster CRC64 and LZF paths for RDB load, a fix for the BITOP lookup regression, and macOS module builds now link against the selected SDK.
This week in Redis: bitmaps can use a native Roaring bitmap representation backed by vendored CROaring. A new INFO Streams section reports stream and consumer-group histograms. CRC64 checksumming gains a PCLMULQDQ path and LZF decompression a faster decoder, both aimed at RDB load. BITOP looks up each source key once again, fixing a regression. Stream MINID approximate trims propagate correctly with tombstones. macOS module builds now resolve the selected SDK.
Week of October 5, 2026
What shipped
- Fixed server crashes when a module thread pauses IO threads with io-threads greater than one. Pull request #15869
- Fixed legacy AOF preload data loss caused by premature upgrade when preload-file points to the local legacy AOF and appendonly is enabled. Pull request #15916
- Exposed phase, snapshot progress percentage and incremental progress percentage in CLUSTER MIGRATION STATUS. Pull request #15829
- Fixed replica node closing its connection because of a full compression input buffer under heavy traffic. Pull request #15912
- Fixed a Linux TCP connect error queue issue in hiredis that could send Sentinel into a busy loop. Pull request #15917
- Added NULL key checks to RM_ListInsert and RM_ListSet so they fail with errno EINVAL as documented. Pull request #15914
- Rejected registering functions with names that differ only by case in the same Lua library. Pull request #15884
- Bounded serialized max links in Vector Sets by the index M to reject crafted payloads. Pull request #15906
- Fixed a null pointer dereference in hnsw reconnect nodes when an allocation fails. Pull request #15911
- Refactored set encoding logic in t set dot c to dispatch through set type operations. Pull request #15766
Why it matters
This week brought several stability fixes that matter for production deployments. Crashes with module threads and IO threads, replica disconnections under heavy traffic, and legacy AOF preload data loss were all addressed. Operators also gained more visibility into cluster migrations with progress percentages in CLUSTER MIGRATION STATUS.
Changelog entry
- Fixed crashes when a module thread pauses IO threads with io-threads greater than one. Pull request #15869
- Fixed legacy AOF preload data loss caused by premature upgrade when preload-file points to the local legacy AOF and appendonly is enabled. Pull request #15916
- Added phase,
snapshot_progress_percandincremental_progress_percto CLUSTER MIGRATION STATUS. Pull request #15829 - Fixed replica node using compression closing the primary client because of a full input buffer. Pull request #15912
- Fixed Linux TCP connect error queue in hiredis to prevent Sentinel busy loop. Pull request #15917
- Added NULL key checks to RM_ListInsert() and RM_ListSet(). Pull request #15914
- Rejected registering functions with names that are the same modulo case in Lua scripts. Pull request #15884
- Bounded serialized max links by the index M in Vector Sets. Pull request #15906
- Fixed null pointer dereference in hnsw_reconnect_nodes() on allocation failure. Pull request #15911
- Refactored
t_set.cto dispatch per-encoding logic via setTypeOps. Pull request #15766 - Skipped the HELLO 2 reconnect test of redis-cli under --force-resp3. Pull request #15888
Fixed: module thread IO thread crashes, legacy AOF preload data loss, replica compression buffer disconnect, hiredis Sentinel busy loop. Added: CLUSTER MIGRATION STATUS progress percentages, NULL key checks for list APIs. Refactored: set encoding dispatch, case-insensitive functi
This week we shipped several stability fixes and improvements. We fixed server crashes when a module thread pauses IO threads, legacy AOF preload data loss caused by premature upgrade, replica disconnections from a full compression input buffer, and a Linux TCP connect error queue issue in hiredis. We also added NULL key checks to RM_ListInsert and RM_ListSet, rejected registering functions with names differing only by case, and bounded serialized max links in Vector Sets by the index M. CLUSTER MIGRATION STATUS now exposes phase, snapshot progress percentage and incremental progress percentage. Internally, set encoding logic was refactored to dispatch through set type operations.
Week of September 28, 2026
What shipped
- TLS builds now work against OpenSSL 4 by replacing the deprecated X509_NAME_get_text_by_NID API. Pull request #15227
- RESP3 protocol preference is now preserved across reconnects and cluster redirects instead of silently falling back to RESP2. Pull request #15806
- Blocking timeouts and module timers now use monotonic time to prevent race conditions where timeouts could be affected by system clock changes. Pull request #15743
- BLMOVEM with EXACTLY modifier now correctly wakes when a module grows an existing list, fixing cases where clients remained blocked unnecessarily. Pull request #15876
- ARLASTITEMS no longer drops elements past holes in the array, properly returning all requested elements. Pull request #15875
- ZSCAN now avoids allocating temporary copies of members and scores, reducing memory churn per scanned member. Pull request #15822
- LoongArch64 architecture now has processor clock support using rdtime.d and cpucfg instructions. Pull request #15769
- Graceful TLS client shutdown no longer produces spurious error logs on the server. Pull request #15758
- Test framework now explains what timed out instead of silently killing hung tests, improving visibility into test failures. Pull request #15879
- GitHub release tarballs now include version tags in their names for better identification. Pull request #15877
Why it matters
This week addresses protocol reliability with RESP3 persistence across reconnects, fixes correctness issues in blocking commands and array operations, and improves compatibility with OpenSSL 4 and new hardware architectures. Test infrastructure improvements make debugging test hangs easier.
Changelog entry
- Fixed: TLS builds against OpenSSL 4 by replacing deprecated X509_NAME_get_text_by_NID API Pull request #15227
- Fixed: RESP3 protocol preference now persists across reconnects and cluster redirects Pull request #15806
- Fixed: Blocking timeouts and module timers now use monotonic time instead of system clock Pull request #15743
- Fixed: BLMOVEM EXACTLY now correctly wakes when modules grow existing lists Pull request #15876
- Fixed: ARLASTITEMS no longer drops elements past holes in the array Pull request #15875
- Changed: ZSCAN reduces memory allocations by avoiding temporary member and score copies Pull request #15822
- Added: Processor clock support for LoongArch64 architecture Pull request #15769
- Fixed: Graceful TLS client shutdown no longer produces spurious error logs Pull request #15758
- Changed: Test framework now explains timed out tests instead of silently killing them Pull request #15879
- Fixed: RM_CreateKeyMetaClass() API documentation corrected for metadata name and version ranges Pull request #15853
Redis releases improvements to RESP3 persistence across reconnects, fixes for blocking commands and array operations, OpenSSL 4 compatibility, and better test diagnostics.
This week's Redis updates focus on protocol reliability and correctness: RESP3 preferences now survive reconnects and cluster redirects, blocking commands wake properly when lists grow, and array operations no longer silently drop elements. The fixes also extend compatibility to OpenSSL 4 and LoongArch64 systems, while test infrastructure improvements provide better visibility into hung tests.
Week of September 21, 2026
What shipped
- Cluster tests migrated from legacy framework to normal test framework, moving 28 test files into the standard test structure. Pull request #15399
- Fixed XTRIM leaving stream nodes with no live entries, which prevented RDB from loading during replication and persistence. Pull request #15815
- RDB and AOF file loading now hint sequential access patterns to the kernel, doubling read-ahead window for faster loads on high-latency storage. Pull request #15782
- Protocol errors are now properly reported when a command spans multiple read calls instead of being silently dropped. Pull request #15795
- Fixed crash during shutdown when stopping AOF in certain states by preventing fsync and close calls on unopened file descriptors. Pull request #15827
- Sentinel no longer enters 100% CPU busy-loop when encountering async socket errors by properly consuming SO_ERROR from the kernel. Pull request #15107
- Fixed race conditions in AOF integration tests caused by multiple servers appending to a shared log file. Pull request #15854
- Restored development version metadata in unstable branch that was accidentally overwritten with RC2 release information. Pull request #15833
- Removed unused
redis_docs_syncCI workflow that was consistently failing. Pull request #15816
Why it matters
This week addresses several stability issues: stream operations now work correctly with persistence, protocol handling is more robust across fragmented reads, and Sentinel stops spinning in tight loops. Infrastructure improvements speed up RDB/AOF loading on cloud storage, and test framework consolidation reduces technical debt.
Changelog entry
- Cluster tests migrated to standard test framework Pull request #15399
- Fix XTRIM leaving 0-live stream nodes that RDB cannot load Pull request #15815
- Hint sequential access when loading RDB and AOF files Pull request #15782
- Protocol errors now reported when command is split across reads Pull request #15795
- Fix crash when closing unopened AOF file during shutdown Pull request #15827
- Fix Sentinel 100% CPU busy-loop on async socket errors Pull request #15107
- Fix log-wait races in AOF integration tests Pull request #15854
Weekly update: fixed XTRIM leaving unpersistable streams, improved protocol error reporting, optimized file loading for cloud storage, and eliminated a Sentinel CPU busy-loop.
This week's improvements focus on stability and performance: XTRIM now works correctly with persistence, protocol errors are properly reported across fragmented reads, and RDB/AOF loading hints sequential access for faster startup on cloud storage. We also fixed a Sentinel CPU busy-loop issue and consolidated cluster tests into the standard framework.
Week of September 14, 2026
What shipped
- Redis 8.10.2 released with security fixes for ACL key permission checks in transactions and cluster bus authentication. Release
- Redis 8.8.3 released with the same security fixes for ACL and cluster bus issues. Release
- Redis 8.6.7 released with security fixes for ACL key permissions and cluster bus authentication. Release
- Redis 8.4.7 released addressing ACL and cluster bus security concerns. Release
- Redis 8.2.10 released with a security fix requiring explicit configuration to use unauthenticated cluster bus ports. Release
- Added BLESS command to protect keys from eviction, with protection surviving restarts and replication. Pull request #15649
- Redis-cli now uses CLIENT NO-TOUCH for --bigkeys, --memkeys, and --keystats to prevent affecting LRU/LFU eviction policy metrics. Pull request #15817
- AOF replay-time estimate now exposed in INFO persistence to help operators gauge reload RTO. Release
- Fixed ASM tests to work with --single flag by ensuring proper slot setup across nodes. Pull request #15636
- Fixed race condition in AOF duration test module where arguments were released before slowlog recording completed. Pull request #15805
Why it matters
This week focused on security patches across multiple release branches addressing critical ACL and cluster authentication issues, along with operational improvements for eviction protection and AOF visibility. Several test reliability fixes and module version updates ensure better stability and observability for production deployments.
Changelog entry
- Security: Fix ACL key permissions not being rechecked at EXEC time Pull request #15804
- Security: Make unauthenticated cluster bus port an explicit choice Pull request #15804
- Feature: Add BLESS command to protect keys from eviction Pull request #15649
- Feature: Expose AOF command duration estimate in INFO persistence Release
- Enhancement: redis-cli uses CLIENT NO-TOUCH for --bigkeys, --memkeys, --keystats Pull request #15817
- Fix: ASM tests now work with --single flag Pull request #15636
- Fix: Race condition in AOF duration test module argument release Pull request #15805
- Fix: AOF duration test tolerance for macOS sleep delay Pull request #15814
- Cleanup: Remove unused EB_BUCKET_KEY_PRECISION machinery from ebuckets Pull request #15725
- Test: Add regression test for loading cluster config without shard IDs Pull request #15757
Redis 8.10.2, 8.8.3, 8.6.7, 8.4.7, and 8.2.10 released with security fixes for ACL key permissions in transactions and cluster bus authentication. BLESS command added to protect keys from eviction.
This week's Redis releases (8.10.2, 8.8.3, 8.6.7, 8.4.7, 8.2.10) address critical security issues with ACL key permission rechecking in transactions and cluster bus authentication. Additionally, the new BLESS command lets operators protect specific keys from eviction, surviving restarts and replication. redis-cli improvements prevent scan operations from affecting eviction metrics, and AOF replay-time estimates now appear in INFO for RTO visibility.