[dependabot]

Bumps the compatible-go-dependencies group with 9 updates in the / directory:

Package	From	To
modernc.org/sqlite	1.58.0	1.60.1
github.com/nats-io/nats.go	1.53.1	1.54.0
github.com/rabbitmq/amqp091-go	1.14.0	1.15.0
github.com/redis/go-redis/v9	9.22.0	9.23.0
github.com/aws/aws-sdk-go-v2	1.46.0	1.47.3
github.com/aws/aws-sdk-go-v2/config	1.33.3	1.33.9
github.com/aws/aws-sdk-go-v2/credentials	1.20.3	1.20.9
github.com/aws/aws-sdk-go-v2/service/sqs	1.51.0	1.52.4
github.com/jackc/pgx/v5	5.10.0	5.11.0

Updates modernc.org/sqlite from 1.58.0 to 1.60.1

Changelog

Sourced from modernc.org/sqlite's changelog.

Changelog

Entries for v1.38.1 through v1.44.1 and for v1.49.1 were added on 2026-09-05, reconstructed from the git history and the merge requests they cite; they were missing at release time.

• 2026-09-30 v1.61.0:

  • Add _time_format=string_no_monotonic, opt-in: times are written in the default layout but without Go's monotonic clock reading, the m=+... suffix, so a scanned time matches its row again when bound in WHERE col = ?. The default is unchanged. Resolves [GitLab issue #259](https://gitlab.com/cznic/sqlite/-/issues/259). See [GitLab merge request #142](https://gitlab.com/cznic/sqlite/-/merge_requests/142), thanks 不插电 (@​r27153733)!
  • Each release tag now gets a signed provenance document, linked from its GitLab release; VERIFYING.md says what it proves and how to check one. Tooling only. See [GitLab merge request #141](https://gitlab.com/cznic/sqlite/-/merge_requests/141).
  • Add an opt-in SQLite authorizer per connection: AuthorizerRegisterer, reached through sql.Conn.Raw or installed on every pooled connection from RegisterConnectionHook; existing connections are unchanged. Resolves [GitLab issue #256](https://gitlab.com/cznic/sqlite/-/issues/256). See [GitLab merge request #139](https://gitlab.com/cznic/sqlite/-/merge_requests/139), thanks Tedla Brandsema (@​TedlaBrandsema)!
  • Closing a connection now forgets the pre-update, commit and rollback hooks registered on it; their entries used to linger after Close.
  • With a shared cache, BeginTx now waits for a lock another connection holds instead of failing at once with SQLITE_LOCKED_SHAREDCACHE (262), as executing BEGIN IMMEDIATE did; Commit and Rollback wait likewise while another connection changes the schema. busy_timeout does not bound the wait; the call's context does. See [GitLab merge request #143](https://gitlab.com/cznic/sqlite/-/merge_requests/143), thanks Ian Chechin!
  • PrepareContext and BeginTx with a context that is already done now return the context's error, as ExecContext and QueryContext do. See [GitLab merge request #143](https://gitlab.com/cznic/sqlite/-/merge_requests/143), thanks Ian Chechin!

• 2026-09-29 v1.60.1:

  • Binding arguments to a statement is no longer quadratic in the number of its parameters, which made multi-row INSERTs with thousands of ? parameters slow. Resolves [GitHub issue #8](modernc-org/sqlite#8), thanks wencycool!

• 2026-09-28 v1.60.0:

  • A fault while reading the memory-mapped -shm file of a WAL database no longer crashes the process. The statement fails with a disk I/O error, extended code SQLITE_IOERR_IN_PAGE (8714), and the connection stays usable, as in MSVC builds of SQLite. On by default on every platform and not switchable; lib.SehInject and lib.SehPending inject such a fault for tests. Resolves [GitLab issue #221](https://gitlab.com/cznic/sqlite/-/issues/221), thanks Roman (@​requilence) for the report, and supersedes libsqlite3!4 and [GitHub pull request #7](modernc-org/sqlite#7), thanks hazyhaar for the two rounds, the ccgo finding and the Windows Server runs!
  • Re-vendor lib/ from modernc.org/libsqlite3 v1.15.0 and vec/ from modernc.org/libsqlite_vec v0.6.0; SQLite stays 3.53.4 and sqlite-vec v0.1.9. Go 1.26 is now required, and the pinned modernc.org/libc becomes v1.77.1; as always, downstream go.mod files must pin the same modernc.org/libc version this repository's go.mod does, see [GitLab issue #177](https://gitlab.com/cznic/sqlite/-/issues/177).
  • Nine exported lib constants that never had a meaningful value are gone: INFINITY, MB_CUR_MAX, NAN, RESERVED_BYTE, SHARED_FIRST, SQLITE_CANTOPEN_BKPT, SQLITE_CORRUPT_BKPT, SQLITE_DEFAULT_LOOKASIDE and SQLITE_MISUSE_BKPT. The transpiler now evaluates object-like macros as C expressions, so other constants take their C value (lib.WALINDEX_PGSZ is 32768, not 0) and some appear; the full list is in libsqlite3's CHANGELOG under 2026-09-19.
  • make vendor now writes vendor.json, the modernc.org/libsqlite3 and modernc.org/libsqlite_vec commits and the Go toolchain lib/ and vec/ were vendored with, so git show vX.Y.Z:vendor.json says which revisions a release carries; the test suite fails when they no longer match. Tooling only. See [GitLab merge request #140](https://gitlab.com/cznic/sqlite/-/merge_requests/140).
  • vfs.FS.Close now refuses while a database opened through it is still open, returning an error that wraps the new vfs.ErrInUse and leaving the VFS registered. It used to free the VFS the open connection still called through, so the next query crashed the process or read through freed memory. Close the databases first, then the FS.
  • Fix handle reuse in modernc.org/sqlite/vfs on 32-bit targets: after 2^32 file opens in one process the handle counter wrapped and could overwrite a live entry, such as a file system registered at start-up, and crash. 64-bit targets were not affected.
  • The pluggable page cache now panics when a Cache breaks its contract by returning nil, or a different Page, from Fetch for a page SQLite still holds pinned. It used to free memory SQLite was still using, corrupting the database without an error. Only a Cache implementation with that bug is affected; modernc.org/sqlite/pcache is not.
  • Document three limits of the pluggable page cache on RegisterPageCache and Cache: it cannot be combined with modernc.org/sqlite/vec, PageCache.Create may be called concurrently, and under cache=shared a Cache is called from several goroutines. Documentation only.
  • Document in the package documentation that state set on a pooled connection is inherited by the next caller to borrow it, and that a connection reached through sql.Conn.Raw is not safe for concurrent use. Documentation only.
  • Add StrictPragmas, opt-in and off by default: once enabled, a connection whose _pragma DSN value holds more than one SQL statement fails to open with ErrMultiStatementPragma, before any DSN parameter is applied. A _pragma value runs as SQL text, so anything after a ; runs too. Recommended for any application whose DSN is not a compile-time constant.
  • Document SQLite's own URI query parameters on Driver.Open: mode, cache, immutable, nolock, psow and modeof, which have always worked in a DSN starting with file:, and the trap that a plain file name has its query stripped before SQLite sees it, so /path/to.db?mode=ro opens read-write. Resolves [GitLab issue #257](https://gitlab.com/cznic/sqlite/-/issues/257). Documentation only.
  • Add SECURITY.md, the vulnerability reporting policy: three private channels, what is in scope, that only the latest release is supported, and how a confirmed report is disclosed, including an entry in the Go vulnerability database so govulncheck reports it. IRP.md, linked from it, is the maintainers' incident response plan. Documentation only.
  • Add CONTRIBUTING.md: where to send a merge request, which files are generated and must not be edited by hand, how to build and test across the 20 supported targets, and the AUTHORS/CONTRIBUTORS convention. Documentation only.
  • Ship LICENSE-3RD-PARTY.md, a transitively flattened inventory of every third-party component this module carries with all seventeen license texts in full, and a Software Bill of Materials, sbom.cdx.json (CycloneDX 1.6) and sbom.spdx.json (SPDX 2.3), explained in SBOM.md. Both cover what a module-graph tool cannot see: the transpiled SQLite and sqlite-vec C, and the upstreams modernc.org/libc carries, musl among them. Documentation only.

• 2026-09-15 v1.59.0:

  • Bump the pinned modernc.org/libc to v1.75.7 and re-vendor lib/ and vec/. The transpiled SQLite is unchanged, still 3.53.4. On the Linux targets the new libc replaces transpiled musl memcpy, memmove, memset, memcmp and strlen with native Go, cutting CPU time on query-heavy workloads by up to a third; see the new Performance section below. As always, downstream go.mod files must pin the same modernc.org/libc version this repository's go.mod does; see [GitLab issue #177](https://gitlab.com/cznic/sqlite/-/issues/177).
  • Hand user-defined function and aggregate callbacks a pooled *FunctionContext instead of allocating a fresh one per call, removing the last driver-side allocation per invocation. Like the argument slice, it is valid only for the duration of the callback and must not be retained past its return. Updates [GitLab issue #226](https://gitlab.com/cznic/sqlite/-/issues/226). See [GitLab merge request #137](https://gitlab.com/cznic/sqlite/-/merge_requests/137).
  • Add regression tests pinning the identity and the pooling of that context. See [GitLab merge request #138](https://gitlab.com/cznic/sqlite/-/merge_requests/138), thanks Ian Chechin!
  • Add a Performance section to the package documentation: measured CPU-time ratios of this driver against the same SQLite compiled from C, where the gap comes from, and the two consequences for applications — index the columns that ORDER BY, GROUP BY and WHERE use, and bound the database/sql pool with SetMaxOpenConns.

• 2026-09-01 v1.58.0:

  • Upgrade to SQLite 3.53.4. It carries upstream's own fix for the journal-rollback data-corruption bug, so the local super-journal patch v1.56.0 introduced is dropped; recovery behavior is unchanged. Also bumps the pinned modernc.org/libc to v1.75.6; as always, downstream modules must pin the same version this one does, see [GitLab issue #177](https://gitlab.com/cznic/sqlite/-/issues/177).
  • Add opt-in support for Linux Open File Description (OFD) locks on database files, off by default; without opting in, locking behavior is byte-for-byte that of previous releases. A POSIX record lock is owned by the (process, inode) pair, so any Close of any descriptor of the database file anywhere in the process silently strips SQLite's locks; OFD locks survive that. Enable it process-wide with MODERNC_SQLITE_OFD_LOCK=1 in the environment, or with the new OFDLocking(true) before the first connection is opened; OFDLockingEnabled reports the mode in effect, and the new ErrOFDLockingTooLate and ErrOFDLockingUnavailable report a switch attempted too late and a platform or filesystem without the feature. Why it is process-wide rather than a DSN parameter, what WAL's -shm coordination still uses, and the /proc/locks measurements behind the design are in [GitLab issue #255](https://gitlab.com/cznic/sqlite/-/issues/255).
  • Resolves [GitLab issue #255](https://gitlab.com/cznic/sqlite/-/issues/255). See [GitLab merge request #136](https://gitlab.com/cznic/sqlite/-/merge_requests/136), thanks Nathan Herring (@​technosloth), and thanks Gani Georgiev (@​ganigeorgiev) for pressing the opt-in default!

• 2026-08-19 v1.57.0:

  • Add an opt-in _defensive DSN query parameter turning on SQLite's defensive mode for the connection. On such a connection PRAGMA writable_schema=ON, PRAGMA journal_mode=OFF and PRAGMA schema_version=N become silent no-ops, and writes to a virtual table's shadow tables and to sqlite_dbpage fail. It is a hardening measure, not a sandbox for hostile database files, for which it is only one of the steps SQLite recommends, and it is a property of the connection, not of the file. Absent, or _defensive=0, nothing changes.
  • Reject _defensive=1 together with _journal_mode=OFF (or _journal=OFF) instead of opening a connection in which neither was honoured: SQLite turns that PRAGMA into a no-op that still reports success. Only DSNs using the new parameter can be affected. See [GitHub pull request #6](modernc-org/sqlite#6), thanks wsman!
  • Ship the sqlite-vec license notice this module has been missing since vec/ arrived in v1.47.0. sqlite-vec is Copyright (c) 2024 Alex Garcia, dual-licensed Apache-2.0 OR MIT and used here under MIT; the text now ships as LICENSE-SQLITE_VEC, and make vendor fails rather than quietly dropping it.
  • The SQLite notice is renamed from SQLITE-LICENSE to LICENSE-SQLITE; update any direct links to it. Its contents are unchanged. The rename is what makes go mod vendor carry both notices into downstream vendor/ trees: it selects license files by name prefix, so a name merely ending in LICENSE was never propagated.
  • Let a caller-constructed Driver register its own functions, collations and virtual table modules, through new RegisterFunction, RegisterScalarFunction, RegisterDeterministicScalarFunction, RegisterCollationUtf8 and RegisterModule methods plus Must* variants, and let vtab.RegisterModule honour its db argument. Behavior change: vtab.RegisterModule(db, ...) where db was opened on a caller-constructed Driver used to discard db and land on the registered sqlite driver, reaching every connection in the process; it now lands on that Driver alone, so a sql.Open("sqlite") connection that used to resolve such a module gets no such module. Everything else is additive, and the isolating change discussed in [GitLab issue #254](https://gitlab.com/cznic/sqlite/-/issues/254) is deliberately not made here. See [GitLab merge request #135](https://gitlab.com/cznic/sqlite/-/merge_requests/135), thanks Ian Chechin!
  • Promote freebsd/386, freebsd/arm and netbsd/amd64 from experimental to fully supported. The package documentation's platform table had carried seventeen entries while this module shipped, cross-built and tested twenty; all three have been in the builder matrix since v1.53.0 and pass the full suite on this release's commit. Documentation only — lib/ is byte-for-byte what v1.56.0 shipped.

... (truncated)

Commits

• b122d04 conn: bind statement arguments in linear time
• 7d5a376 CHANGELOG.md: slim the v1.60.0 section
• a604165 CHANGELOG.md: the SEH emulation, the re-vendor and the removed lib constants;...
• fb4aac4 vendor.json: stamp lib/ from libsqlite3 v1.15.0 and vec/ from libsqlite_vec v...
• 0f7ae7d Merge branch 'master' into seh-emulation
• 2e43324 lib: re-vendor from modernc.org/libsqlite3 v1.15.0 (SEH emulation), pin libc ...
• 50380ee lib, tests: Go side of the wal.c SEH emulation for Windows in-page faults (#221)
• 4552e53 Merge branch 'vendor-stamp' into 'master'
• 3775177 Makefile, doc.go, IRP.md: VENDORFLAGS for debug builds; name every refusal
• 86c97d4 CHANGELOG.md: link merge request !140
• Additional commits viewable in compare view

Updates github.com/nats-io/nats.go from 1.53.1 to 1.54.0

Release notes

Sourced from github.com/nats-io/nats.go's releases.

Release v1.54.0

Changelog

This release raises the minimum supported Go version to 1.26 and adds support for nats-server v2.15 stream and consumer info fields.

ADDED

• Core NATS:
  • nats.MultipathTCP() option to enable or disable MPTCP for outbound connections (#2145)
  • Conn.ConnectedDomain() to read the JetStream domain advertised by the connected server (#2139)
• JetStream:
  • ClusterInfo.Desired, PeerInfo.Peer, PeerInfo.Pending and StreamSourceInfo.Seq in stream and consumer info, reported by nats-server v2.15 (#2144)
  • Nats-Schedule-Rollup header const and WithScheduleRollup() publish option (#2142)

FIXED

• Core NATS:
  • Panic in DecodeHeadersMsg on inline status tokens shorter than 3 characters. Thanks @​nikitha-m for the contribution (#2101)
  • Data race on Subscription.sid between applyNewSID and removeSub (#2122)
• JetStream:
  • Messages() iterator going silent after a reconnect that completes between Next() calls. Thanks @​1995parham for the contribution (#2135)
  • Legacy ordered consumer silently losing messages when the subscription hits its pending limits (#2137)
  • Legacy ordered consumer reset resurrecting an unsubscribed or draining subscription (#2133)

IMPROVED

• Core NATS:
  • Avoid allocation when parsing header message arguments with a fifth token. Thanks @​0xAozora for the contribution (#2130)
  • Preallocate reply subject buffers. Thanks @​0xAozora for the contribution (#2123)
• JetStream:
  • Prevent doErr closure escaping to the heap in async publish reply handling. Thanks @​0xAozora for the contribution (#2127)
  • Fix typo in Publisher interface methods comment. Thanks @​sreeram77 for the contribution (#2120)
• Minimum Go version is now 1.26 (#2146)
• Bump klauspost/compress to v1.20.0 and nkeys to v0.4.16 (#2117, #2146)
• Tests run locally without docker via an in-process ntf tester (#2131), and CI now runs against nats-server v2.15.0 (#2146)
• Fix flaky cluster restart tests (#2141) and list consumers test against latest server (#2138)

Complete Changes

nats-io/nats.go@v1.53.1...v1.54.0

Commits

• 7a8404a Release v1.54.0 (#2148)
• 240fb1f [IMPROVED] Bump dependencies and update server version in CI (#2146)
• 203d828 [FIXED] Panic in DecodeHeadersMsg on status tokens shorter than 3 characters ...
• dc92fca [FIXED] Ordered consumer silently losing messages on slow consumer (#2137)
• e18cea8 [IMPROVED] Prevent doErr closure escaping to heap (#2127)
• d8d8c09 [ADDED] nats.MultipathTCP() for enabling or disabling MPTCP (#2145)
• e95c424 [FIXED] Flaky cluster restart tests: waitForStream must see a stream leader (...
• e3abbf4 [ADDED] Nats-Schedule-Rollup header const and publish options (#2142)
• f1289db [ADDED] Desired state, peer ID/pending and source seq in stream and consumer ...
• b84f783 Ignore Claude Code local files and .worktrees (#2140)
• Additional commits viewable in compare view

Updates github.com/rabbitmq/amqp091-go from 1.14.0 to 1.15.0

Release notes

Sourced from github.com/rabbitmq/amqp091-go's releases.

v1.15.0

What's Changed

• chore(deps): bump github/codeql-action from 4.37.6 to 4.37.7 in the github-actions group by @​dependabot[bot] in rabbitmq/amqp091-go#385
• chore(deps): bump github/codeql-action from 4.37.7 to 4.37.8 in the github-actions group by @​dependabot[bot] in rabbitmq/amqp091-go#386
• fix: clone SASL authentication before zeroing credentials post-handshake by @​suchitd in rabbitmq/amqp091-go#388
• fix: preserve caller-provided SASL credentials across Reconnect() by @​suchitd in rabbitmq/amqp091-go#389
• chore(deps): bump github/codeql-action from 4.37.8 to 4.37.9 in the github-actions group by @​dependabot[bot] in rabbitmq/amqp091-go#390

Full Changelog: rabbitmq/amqp091-go@v1.14.0...v1.15.0

Changelog

Sourced from github.com/rabbitmq/amqp091-go's changelog.

v1.15.0 (2026-09-15)

Full Changelog

Fixed bugs:

• fix: preserve caller-provided SASL credentials across Reconnect() #389 (suchitd)
• fix: clone SASL authentication before zeroing credentials post-handshake #388 (suchitd)

Closed issues:

• v1.13.0 (#350) zeroes caller-provided PlainAuth.Password, breaking auth on any reconnect that reuses the Config #387

Merged pull requests:

• chore(deps): bump github/codeql-action from 4.37.8 to 4.37.9 in the github-actions group #390 (dependabot[bot])
• chore(deps): bump github/codeql-action from 4.37.7 to 4.37.8 in the github-actions group #386 (dependabot[bot])
• chore(deps): bump github/codeql-action from 4.37.6 to 4.37.7 in the github-actions group #385 (dependabot[bot])

Commits

• ff03b2f Release v1.15.0
• 276f52e Merge pull request #390 from rabbitmq/dependabot/github_actions/github-action...
• ca02321 chore(deps): bump github/codeql-action in the github-actions group
• 4a2300f Merge pull request #389 from rabbitmq/fix/recovery-preserve-sasl
• bf771db docs: trim cloneAuthentication doc comment
• 54bd90e fix: preserve caller-provided SASL credentials across Reconnect()
• 5935ca1 fix: clone SASL authentication before zeroing credentials post-handshake (#388)
• 0e52b40 Merge pull request #386 from rabbitmq/dependabot/github_actions/github-action...
• 01c30ed chore(deps): bump github/codeql-action in the github-actions group
• 28da80d Merge pull request #385 from rabbitmq/dependabot/github_actions/github-action...
• Additional commits viewable in compare view

Updates github.com/redis/go-redis/v9 from 9.22.0 to 9.23.0

Release notes

Sourced from github.com/redis/go-redis/v9's releases.

9.23.0

This is a minor release. It contains everything from 9.23.0-beta.1, so these notes cover the full 9.22.0 → 9.23.0 upgrade.

⚠️ Changes to check when you upgrade from 9.22.0:

• Go 1.26 is now the minimum Go version (#4060). The go directive of the root module and of the submodules moved from 1.24 to 1.26, as part of the dependency bumps for govulncheck findings. Projects that build with Go 1.24 or 1.25 must upgrade the toolchain.
• Each Client has a dedicated pipeline connection pool (#4002). This includes failover clients and cluster node clients. Pipeline, TxPipeline, and autopipeline operations use this pool, so they do not compete with regular commands for main-pool connections. The pool is for burst capacity only: it dials only when necessary, an unused pool holds zero connections, and if the pool is full an operation uses the main pool. Set PipelinePoolSize: -1 to get the previous single-pool behavior.
• AutoPipelineOptions.MaxBatchBytes has a default of 128 KiB (before, there was no limit). The default prevents a write/reply deadlock; it is not a throughput control. The buffers of the pipeline pool also have a default of 128 KiB.

⚠️ Changes to experimental APIs since 9.23.0-beta.1 (no effect if you upgrade from 9.22.0):

• AutoPipelineOptions.FullDuplexFastSubmit is removed (#4018). The submit channel is replaced by a queue that takes a full wave of commands in one lock, so the separate fast path is not necessary. Remove the field from your options.
• Options.ClientSideCacheRefreshRecencyWindow is removed (#4034). Refresh-on-invalidate now reads again every cached entry of an invalidated key and does not look at how recently the entry was read. On a write-heavy keyspace this means refresh traffic across the full resident cache. Remove the field from your options.
• LocalCache.LRUClock() is removed (#4034). It returned the recency token that the removed recency window used, and has no replacement. Remove calls to it.
• ClientSideCacheInvalidationBatchWindow served stale values in beta.1 (#4033). Under load the batcher applied only ~15% of invalidations, and the cache served invalidated values with a ~100% hit rate. 9.23.0 applies all of them. If you used this option on beta.1, upgrade.

🚀 Highlights

Full-Duplex Auto-Pipelining (Experimental)

Set AutoPipelineOptions.FullDuplex to enable the full-duplex mode of the automatic pipeliner. The default mode sends one batch per round trip. In full-duplex mode the engine holds one pipeline-pool connection, and a writer goroutine and a reader goroutine move the ordered command stream in both directions at the same time. On a high-latency link each command completes in approximately one round-trip time (RTT) on a single connection. On a 50 ms WAN profile: ~389k ops/s at 52 ms p50, against ~207k ops/s at 116 ms on the half-duplex ordered path.

The mode works on both faces of a standalone Client (AutoPipeline() and AsyncAutoPipeline()), and natively on a ClusterClient when routing goes to masters only. On a cluster, one child engine per master sends each command to the node that owns its slot, and follows MOVED / ASK redirects and retryable replies (LOADING, READONLY, ...) through the usual cluster redirect path. With replica routing (ReadOnly, RouteByLatency, or RouteRandomly), the autopipeliner uses the half-duplex shard flushers, which obey the shard picker. Config().FullDuplex reports the mode in effect.

Options:

• FullDuplexWindow — the maximum number of commands in flight (backpressure).
• FullDuplexIdleTimeout and FullDuplexMaxHold — when the engine returns the held connection to the pool. The pool hooks then do re-authentication and maintenance-notification handoffs.
• MaxFlushDelay — now also used in full-duplex mode (#4014). The writer waits only when enough commands are in flight, so low-concurrency callers keep the 1×RTT behavior. With MaxFlushDelay=250µs at 1024 callers: +22% ops/s and −26% p99. The default is 0, so this is opt-in.
• NumShards — on a standalone Client, the number of full-duplex engines (#4022). See below.

Pipeline() on the full-duplex connection (#4021). Standalone Client only. Before, ap.Pipeline() used a pooled connection for each Exec. It now submits the batch to the engine as one contiguous run (FDPipelined), so replies come back in submit order, as on a dedicated connection. At 256–1024 callers with 10 commands per Exec: +49% to +52% throughput on one socket instead of ~150. Errors, retries, hooks, and metrics behave like an ordinary pipeline. A batch that cannot use the stream (a blocking command, a per-command read timeout, a diverted command, or a batch larger than the queue) falls back to the ordinary pipeline. TxPipeline stays on a pooled connection, because MULTI/EXEC needs connection affinity. On a ClusterClient, ap.Pipeline() is still the ordinary cluster pipeline. A batch that the queue cannot admit now reserves its slots and waits in a first-come line, so single commands cannot starve a long pipeline (#4057, merged as part of #4021) by @​vlady-kotsev.

Several engines on one client (#4022). With FullDuplex, NumShards > 1 now means N engines on one client: one pool, one hook chain, N held connections. Commands route by a hash of their first key, so commands on the same key stay on one connection and keep their order without Unordered. Keyless commands are distributed round-robin. An FDPipelined batch stays on one engine. The pipeline pool must hold at least NumShards connections for each full-duplex autopipeliner. More engines help only under high concurrency: with 10-command pipelines, 8 engines were 30% slower than 1 at 8 callers, equal at about 128 callers, and 2.2× faster at 1024 callers.

The full-duplex path supports blocking commands, Options.Limiter (one admission per written batch), per-command and pipeline hooks, OTel metrics, retry budgets (a NoRetry command is never sent again after it is on the wire), and seamless maintenance handoffs. Limits of the stream model:

• A hook can observe a command, but cannot stop it. A ProcessHook that returns without calling next does not cancel the command on the full-duplex path; the command is already queued on the held connection. Run policy or kill-switch hooks on a plain client or on the half-duplex autopipeliner.
• Diverted commands are not ordered with the stream. The engine diverts blocking commands, connection-hostile commands, and managed HIMPORT commands. On the async face, wait for the result of a diverted command before you submit a command that depends on it.
• Single commands on the full-duplex stream do not use the client-side cache. If CSC is enabled, a cacheable command on the stream does not read or write the cache.
• Duration metrics are recorded per reply. A command that fails before it reaches the reader (lease failure, limiter denial, retry exhaustion, close) calls the error callback but records no operation-duration sample.

rdb := redis.NewClient(&redis.Options{
	Addr: "localhost:6379",
	AutoPipelineOptions: &redis.AutoPipelineOptions{
		FullDuplex: true, // stream commands on one held connection, ~1 RTT each
	},
})
defer rdb.Close()
</tr></table> 

... (truncated)

Changelog

Sourced from github.com/redis/go-redis/v9's changelog.

9.23.0 (2026-10-05)

This is a minor release. It contains everything from 9.23.0-beta.1, so these notes cover the full 9.22.0 → 9.23.0 upgrade. The release adds:

• An experimental full-duplex mode for the automatic pipeliner, now with Pipeline() on the full-duplex connection and several engines on one client.
• Refresh-on-invalidate and miss coalescing for client-side caching, and a faster cache read path.
• Better distribution of cluster reads under latency-based routing.
• Client.EphemeralConn, an opt-in limit on queued autopipeline commands, and many stability fixes.

⚠️ Changes to check when you upgrade from 9.22.0:

• Go 1.26 is now the minimum Go version (#4060). The go directive of the root module and of the submodules moved from 1.24 to 1.26, as part of the dependency bumps for govulncheck findings. Projects that build with Go 1.24 or 1.25 must upgrade the toolchain.
• Each Client has a dedicated pipeline connection pool (#4002). This includes failover clients and cluster node clients. Pipeline, TxPipeline, and autopipeline operations use this pool, so they do not compete with regular commands for main-pool connections. The pool is for burst capacity only: it dials only when necessary, an unused pool holds zero connections, and if the pool is full an operation uses the main pool. Set PipelinePoolSize: -1 to get the previous single-pool behavior.
• AutoPipelineOptions.MaxBatchBytes has a default of 128 KiB (before, there was no limit). The default prevents a write/reply deadlock; it is not a throughput control. The buffers of the pipeline pool also have a default of 128 KiB.

⚠️ Changes to experimental APIs since 9.23.0-beta.1 (no effect if you upgrade from 9.22.0):

• AutoPipelineOptions.FullDuplexFastSubmit is removed (#4018). The submit channel is replaced by a queue that takes a full wave of commands in one lock, so the separate fast path is not necessary. Remove the field from your options.
• Options.ClientSideCacheRefreshRecencyWindow is removed (#4034). Refresh-on-invalidate now reads again every cached entry of an invalidated key and does not look at how recently the entry was read. On a write-heavy keyspace this means refresh traffic across the full resident cache. Remove the field from your options.
• LocalCache.LRUClock() is removed (#4034). It returned the recency token that the removed recency window used, and has no replacement. Remove calls to it.
• ClientSideCacheInvalidationBatchWindow served stale values in beta.1 (#4033). Under load the batcher applied only ~15% of invalidations, and the cache served invalidated values with a ~100% hit rate. 9.23.0 applies all of them. If you used this option on beta.1, upgrade.

🚀 Highlights

Full-Duplex Auto-Pipelining (Experimental)

Set AutoPipelineOptions.FullDuplex to enable the full-duplex mode of the automatic pipeliner. The default mode sends one batch per round trip. In full-duplex mode the engine holds one pipeline-pool connection, and a writer goroutine and a reader goroutine move the ordered command stream in both directions at the same time. On a high-latency link each command completes in approximately one round-trip time (RTT) on a single connection. On a 50 ms WAN profile: ~389k ops/s at 52 ms p50, against ~207k ops/s at 116 ms on the half-duplex ordered path.

The mode works on both faces of a standalone Client (AutoPipeline() and AsyncAutoPipeline()), and natively on a ClusterClient when routing goes to masters only. On a cluster, one child engine per master sends each command to the node that owns its slot, and follows MOVED / ASK redirects and retryable replies (LOADING, READONLY, ...) through the usual cluster redirect path. With replica routing (ReadOnly, RouteByLatency, or RouteRandomly), the autopipeliner uses the half-duplex shard flushers, which obey the shard picker. Config().FullDuplex reports the mode in effect.

Options:

• FullDuplexWindow — the maximum number of commands in flight (backpressure).
• FullDuplexIdleTimeout and FullDuplexMaxHold — when the engine returns the held connection to the pool. The pool hooks then do re-authentication and maintenance-notification handoffs.
• MaxFlushDelay — now also used in full-duplex mode (#4014). The writer waits only when enough commands are in flight, so low-concurrency callers keep the 1×RTT behavior. With MaxFlushDelay=250µs at 1024 callers: +22% ops/s and −26% p99. The default is 0, so this is opt-in.
• NumShards — on a standalone Client, the number of full-duplex engines (#4022). See below.

Pipeline() on the full-duplex connection (#4021). Standalone Client only. Before, ap.Pipeline() used a pooled connection for each Exec. It now submits the batch to the engine as one contiguous run (FDPipelined), so replies come back in submit order, as on a dedicated connection. At 256–1024 callers with 10 commands per Exec: +49% to +52% throughput on one socket instead of ~150. Errors, retries, hooks, and metrics behave like an ordinary pipeline. A batch that cannot use the stream (a blocking command, a per-command read timeout, a diverted command, or a batch larger than the queue) falls back to the ordinary pipeline. TxPipeline stays on a pooled connection, because MULTI/EXEC needs connection affinity. On a ClusterClient, ap.Pipeline() is still the ordinary cluster pipeline. A batch that the queue cannot admit now reserves its slots and waits in a first-come line, so single commands cannot starve a long pipeline (#4057, merged as part of #4021) by @​vlady-kotsev.

Several engines on one client (#4022). With FullDuplex, NumShards > 1 now means N engines on one client: one pool, one hook chain, N held connections. Commands route by a hash of their first key, so commands on the same key stay on one connection and keep their order without Unordered. Keyless commands are distributed round-robin. An FDPipelined batch stays on one engine. The pipeline pool must hold at least NumShards connections for each full-duplex autopipeliner. More engines help only under high concurrency: with 10-command pipelines, 8 engines were 30% slower than 1 at 8 callers, equal at about 128 callers, and 2.2× faster at 1024 callers.

The full-duplex path supports blocking commands, Options.Limiter (one admission per written batch), per-command and pipeline hooks, OTel metrics, retry budgets (a NoRetry command is never sent again after it is on the wire), and seamless maintenance handoffs. Limits of the stream model:

• A hook can observe a command, but cannot stop it. A ProcessHook that returns without calling next does not cancel the command on the full-duplex path; the command is already queued on the held connection. Run policy or kill-switch hooks on a plain client or on the half-duplex autopipeliner.
• Diverted commands are not ordered with the stream. The engine diverts blocking commands, connection-hostile commands, and managed HIMPORT commands. On the async face, wait for the result of a diverted command before you submit a command that depends on it.
• Single commands on the full-duplex stream do not use the client-side cache. If CSC is enabled, a cacheable command on the stream does not read or write the cache.
• Duration metrics are recorded per reply. A command that fails before it reaches the reader (lease failure, limiter denial, retry exhaustion, close) calls the error callback but records no operation-duration sample.

rdb := redis.NewClient(&redis.Options{
</tr></table> 

... (truncated)

Commits

• 6f2b263 chore(release): v9.23.0 (#4074)
• ec50c69 feat(conn): add DisposableConn to discard on Close (#4039)
• e3d2548 fix(pool): fix ConnStateMachine waiter races (#4072)
• 2ee6d86 fix(pool): wake waiters when a connection is released to the pool (#4028)
• 83d45ed fix:(proto): support uintptr in WriteArg (#4054)
• c2002af docs(example): add full-duplex examples (#4061)
• b5f8864 feat(autopipeline): add MaxQueuedCommands hard limit (#4070)
• 89ecbc5 fix(options): reject invalid skip_verify values in ParseURL (#4064)
• f813adc feat(autopipeline): run N full-duplex engines on one client (#4022)
• 349bcf4 feat(autopipeline): run Pipeline on the full-duplex connection (#4021)
• Additional commits viewable in compare view

Updates github.com/aws/aws-sdk-go-v2 from 1.46.0 to 1.47.3

Commits

• c8f4a44 Release 2026-10-09.2
• 3d21015 update to smithy-go v1.28.5 to pick up sparse-in-dense fix (#3587)
• b058f58 Release 2026-10-09
• 6b5b3dd Regenerated Clients
• 2a845ae Update API model
• c88f24b Release 2026-10-08
• 236b4e5 Regenerated Clients
• 1935912 Update API model
• ab2110b Update to smithy-go v1.28.4 (#3584)
• 313f364 Release 2026-10-06
• Additional commits viewable in compare view

Updates github.com/aws/aws-sdk-go-v2/config from 1.33.3 to 1.33.9

Commits

• c8f4a44 Release 2026-10-09.2
• 3d21015 update to smithy-go v1.28.5 to pick up sparse-in-dense fix (#3587)
• b058f58 Release 2026-10-09
• 6b5b3dd Regenerated Clients
• 2a845ae Update API model
• c88f24b Release 2026-10-08
• 236b4e5 Regenerated Clients
• 1935912 Update API model
• ab2110b Update to smithy-go v1.28.4 (#3584)
• 313f364 Release 2026-10-06
• Additional commits viewable in compare view

Updates github.com/aws/aws-sdk-go-v2/credentials from 1.20.3 to 1.20.9

Commits

• 30383d5 Release 2023-04-07
• 352f89c Regenerated Clients
• 5042939 Update API model
• d40a16e NXDOMAIN errors should not be retried (#2083)
• 439f88c Add announcement for next release for dlm
• e4036a9 Release 2023-04-06
• 1c455e2 Regenerated Clients
• adb09d3 Update endpoints model
• 3d4ed44 Update API model
• 296e005 Release 2023-04-05
• Additional commits viewable in compare view

Updates github.com/aws/aws-sdk-go-v2/service/sqs from 1.51.0 to 1.52.4

Commits

• 7191d24 Release 2024-08-09
• cc3c60f Regenerated Clients
• 50857c3 Update endpoints model
• 50da876 Update API model
• Description has been truncated

[dependabot]

Looks like these dependencies are updatable in another way, so this is no longer needed.