リリース

Flux v2.9.1は、post-build変数置換がFlux自身のCRDスキーマを壊してしまう不具合を修正したパッチリリースです。あわせてSOPSの.ini復号とdry-run時のstrategic merge patchエラーも修正しています。新規の破壊的変更やCVEはありません。

• breaking変数置換によるCRDスキーマ破損を修正

  post-build変数置換を有効にしたKustomizationで、マニフェスト内の${...}がFlux CRDのスキーマフィールドと一致してしまうと、これまでのバージョンではCRDスキーマが壊れることがありました。v2.9.1ではFlux自身のCRDにkustomize.toolkit.fluxcd.io/substitute: disabledを付与し、置換の対象から除外しています。post-build置換を使っている場合はアップグレードを推奨します。

Prometheus v3.13.0はLTSリリースで、HIGH深刻度の認証情報転送の脆弱性(CVE-2025-4673・CVE-2023-45289)とMEDIUM深刻度のXSS(CVE-2026-44990)への対応を筆頭に、ページネーショントークン、パス解決、PromQL期間関数名、ライセンスファイル配布の4つの破壊的変更、さらに多数の新機能とパフォーマンス改善を含みます。

• securityHIGH: クロスホストリダイレクト時に認証情報が転送されなくなった

  リダイレクト先のホストが異なる場合、Authorizationヘッダー・Basic認証・Bearerトークン・OAuth2・カスタムヘッダーなどの認証情報が転送されなくなりました。スクレイプ・リモートread/write・アラート送信・サービスディスカバリのいずれも対象です。CVE-2025-4673とCVE-2023-45289として追跡されており、prometheus/commonのv0.68.xからv0.69.0への更新が起因です。クロスホストリダイレクトに依存するエンドポイントがあれば、認証情報がサイレントに落ちることを確認してください。

• securityMEDIUM: UI依存ライブラリのXSS修正(CVE-2026-44990)

  UIのsanitize-htmlライブラリにXSS脆弱性(CVE-2026-44990)が存在していたため、バージョンを更新して修正しました。設定変更は不要で、v3.13.0へのアップグレードで対処されます。

• breakingページネーショントークンのアルゴリズムがSHA-1からSHA-256に変更

  ルールグループのページネーショントークン生成がSHA-1からSHA-256に変わりました。旧バージョンで取得したトークンを保持・比較しているクライアントやツールは、アップグレード後に値が変わります。

• breaking--http.config.fileの相対パス解決先が変更

  --http.config.fileに渡すファイル内の相対パスが、親ディレクトリではなくそのconfigファイル自身のディレクトリを基準に解決されるようになりました。相対パスを使っている場合は、アップグレード後にパスが正しく解決されるか確認してください。

• breaking期間表現関数min()/max()がmin_of()/max_of()に改名

  experimental-duration-exprフィーチャーフラグを有効にしている場合、PromQLの期間表現関数min()とmax()がmin_of()とmax_of()に改名されました。これらを使用しているクエリやルールを更新してください。

• breakingnpm_licenses.tar.bz2を廃止、ライセンスは/assets/third-party-licenses.txtへ移行

  リリースtarballおよびコンテナイメージからnpm_licenses.tar.bz2が削除されました。サードパーティnpmライセンス情報はバイナリに埋め込まれ、/assets/third-party-licenses.txtで提供されます。このアーカイブを展開している自動化処理があれば修正が必要です。

Longhorn v1.11.3は、ボリューム操作・バックアップ・インスタンス管理にまたがる十数件の安定性問題を修正するバグフィックスロールアップです。CSI external provisionerがv6.3.0へ更新されたため、v1.10.xまたはv1.11.0からアップグレードする場合はKubernetes v1.34以降が前提条件となります。

• breakingv1.10.xまたはv1.11.0からのアップグレード時にKubernetes v1.34+が必要

  CSI external provisionerがv6.3.0に更新されたため、Longhorn v1.10.xまたはv1.11.0からアップグレードする場合はKubernetes v1.34以降が必要です。クラスタのKubernetesバージョンを先にv1.34+へ更新してからLonghornをアップグレードしてください。

wasmCloud v2.5.0 はフィーチャーリリースです。wasmtime 46 への更新と wasip3 のデフォルト有効化、wasmcloud:keyvalue bucket WIT の破壊的変更、および quinn-proto の MEDIUM セキュリティ修正 (RUST-SEC-2026-0185) が主な変更点です。wasmcloud:keyvalue を利用しているコンポーネントは WIT の対応が必要です。

• securityquinn-proto のセキュリティ修正 (RUST-SEC-2026-0185)

  MEDIUM 深刻度の rust-sec-2026-0185 (quinn-proto) を修正しています。QUIC 通信を利用する環境では v2.5.0 へ更新してください。

• breakingwasmcloud:keyvalue bucket のWIT定義が破壊的変更

  wasmcloud:keyvalue の bucket 型がインターフェースごとのインライン定義から共有 types インターフェース経由の定義に移動しました。wasmcloud:keyvalue WIT を利用するコンポーネントやプロバイダは WIT の修正とリビルドが必要です。

• breakingwasip3 がデフォルト有効化

  wasip3 が v2.5.0 からデフォルトで有効になりました。wasip3 対応外のワークロードで予期しない動作が発生する場合は、設定で明示的に無効化してください。

Flux v2.9.0 は機能拡張リリースです。CLI プラグインシステムやセキュリティ機能の充実など複数の新機能を導入する一方、image.toolkit.fluxcd.io/v1beta2 および notification.toolkit.fluxcd.io/v1beta2 の 2 つの API を廃止します。これらの API を使用しているマニフェストがある場合は、アップグレード前に v1 系への移行が必要です。

• breakingimage.toolkit.fluxcd.io/v1beta2 の廃止

  image.toolkit.fluxcd.io/v1beta2 を使用しているマニフェストまたはカスタムリソースは v2.9.0 では動作しません。v1 系 API へ移行してください。

• breakingnotification.toolkit.fluxcd.io/v1beta2 の廃止

  notification.toolkit.fluxcd.io/v1beta2 を使用しているマニフェストまたはカスタムリソースは v2.9.0 では動作しません。v1 系 API へ移行してください。

Kubescape v4.0.10 is a feature and hardening release: it closes an SSRF gap in the scan-completion webhook, tightens config.json permissions and validation, and removes a dead CRD, alongside a large batch of bug fixes and new capabilities like `operator remediate` and encrypted report decryption. No CVE-tracked vulnerabilities are disclosed in this release.

• securitySSRF hardening on scan-completion webhook

  The scan-completion webhook callback now hardens against SSRF. Applies to users of the webhook callback feature; upgrade if you rely on it to avoid outbound requests being abused via crafted callback URLs.

• breakingconfig.json permissions restricted to owner-only

  config.json is now written with owner-only permissions instead of broader access. Applies unconditionally; review any tooling or automation that reads this file as a different user, since it may now fail to access it.

• breakingRemoval of orphan SecurityException CRD

  The orphan SecurityException CRD, which was never installable, has been removed. Applies only if you had references to this CRD in manifests or docs; it has no functional impact since it never worked.

• enhancementStricter validation on --format and VAP controls

  The --format flag now rejects invalid values before a scan starts, and VAP configurable controls now require params. Applies if you script scans with an unchecked --format value or run VAP controls without required params; these calls will now fail fast instead of proceeding silently.

v1.18.1 is a routine patch that fixes CIDR matching behavior and serialization determinism while extending an experimental flag to BatchCheck. No breaking changes or CVEs are reported, though the in_cidr fix does change matching behavior for IPv4-mapped IPv6 addresses.

• breakingin_cidr now matches IPv4-mapped IPv6 addresses against IPv4 CIDRs

  Applies unconditionally: the in_cidr condition now normalizes IPv4-mapped IPv6 addresses (e.g. ::ffff:192.168.1.1) to their IPv4 form before matching, so they can now match IPv4 CIDRs like 192.168.1.0/24. Review any authorization models relying on in_cidr if you expect these addresses to be excluded from IPv4 ranges.

NATS server v2.14.3 is a maintenance release dominated by JetStream, MQTT and clustering bug fixes, with two MQTT fixes closing pre-auth memory exhaustion and panic conditions and one breaking change: JSONP monitoring support has been removed.

• securityMQTT memory exhaustion and panic fixes

  Applies to servers exposing MQTT. Fixed in v2.14.3: malformed partial CONNECT packets could exhaust pre-auth memory, and a PUBLISH remaining-length underflow could panic the server. Upgrade if you run MQTT, especially with untrusted or public clients.

• breakingJSONP monitoring support removed

  Applies to deployments relying on JSONP callbacks against monitoring endpoints (e.g. /varz, /connz via callback= param). This is removed in v2.14.3; switch dashboards or scripts to plain JSON polling before upgrading.

• breakingPermission checks tightened for leaf trace destination and NoAuthUser

  Applies to leaf node setups using Nats-Trace-Dest, and to accounts using NoAuthUser. Fixed in v2.14.3: leaf connections could bypass Nats-Trace-Dest publish permission checks, and NoAuthUser did not enforce connection restrictions. Review permission and NoAuthUser configs after upgrading, since previously permitted traffic may now be denied as intended.

v2.12.12 is a bugfix and hardening release for nats-server, centered on numerous JetStream/Raft data-integrity and panic fixes plus removal of JSONP support from monitoring endpoints. No CVEs are disclosed, but several fixes close memory-exhaustion and panic paths reachable via malformed MQTT input that operators should treat as security-relevant.

• securityMQTT partial-CONNECT and PUBLISH-underflow crash/DoS fixes

  If you run MQTT, upgrade to fix two memory-exhaustion/panic paths: partial CONNECT packets could exhaust pre-authentication memory, and PUBLISH remaining-length underflow could panic the server. Both are exploitable by unauthenticated clients sending malformed input.

• securityInteger-overflow panics fixed in monitoring and JetStream usage tracking

  CONNZ/SUBSZ pagination previously could panic on integer overflow, and JetStream remote usage updates could panic on length integer overflow. Operators with large clusters or heavy monitoring polling should upgrade to remove these panic-inducing edge cases.

• breakingJSONP support removed from monitoring endpoints

  JSONP callback support has been removed from monitoring endpoints in v2.12.12, unconditionally. Any tooling or dashboards that rely on JSONP callbacks against /varz, /connz, or similar endpoints will break and must switch to plain JSON requests.

Strimzi 1.1.0 is a feature release adding Kafka 4.3.0 and 4.2.1 support while dropping Kafka 4.1.x, along with several capability additions. Operators should act on two breaking changes before upgrading: the entity-operator healthcheck port renames and the TLS truststore format switch in KafkaBridge and KafkaMirrorMaker2.

• breakingEntity-operator healthcheck ports renamed

  The entity-operator healthcheck port `healthcheck` has been renamed to `healthcheck-to` (topic-operator) and `healthcheck-uo` (user-operator). Any custom PodMonitor, ServiceMonitor, NetworkPolicy, or similar resources that reference the old port name must be updated before or immediately after upgrading.

• breakingKafka 4.1.x support removed

  Kafka 4.1.x is no longer supported in 1.1.0. Clusters running Kafka 4.1.x must be upgraded to 4.2.1 or 4.3.0 before moving to this Strimzi version.

• breakingTLS truststore format changed to PEM for KafkaBridge and KafkaMirrorMaker2

  KafkaBridge and KafkaMirrorMaker2 now use PEM files instead of P12/JKS for TLS authentication and truststore. Review any tooling or external systems that expect P12/JKS formats from these components.

• breakingCRD v1 only — older API versions unsupported

  CRD API versions v1beta2, v1beta1, and v1alpha1 are not supported since 1.0.0. If you have not already converted your resources to v1, do so before upgrading to 1.1.0.

v1.18.0 is a bugfix rollup with one operator-facing behavior change: the outbound User-Agent header is now hyphenated to conform to RFC 9110. The remaining changes are runtime improvements, formatter and coverage tool fixes, and a Go toolchain bump.

• breakingUser-Agent header format changed (RFC 9110 conformance)

  The outbound User-Agent header OPA sends on HTTP requests changed from 'Open Policy Agent/<version> (<os>, <arch>)' to 'Open-Policy-Agent/<version> (<os>, <arch>)' (hyphen added between 'Open' and 'Policy'). Any server-side log filters, WAF rules, or access policies that exact-match the old string must be updated after upgrading to v1.18.0.

v1.19.6 is a security patch fixing a HIGH severity RBAC privilege escalation in the cert-manager-edit ClusterRole that let users create ACME Challenge and Order resources directly, plus a Go toolchain update closing three CVEs. The RBAC fix includes a documented breaking permission change.

• securityRBAC privilege escalation via ACME Challenge/Order creation

  Applies to v1.19.6: the default cert-manager-edit aggregate ClusterRole previously let namespace users create ACME Challenge and Order resources directly, bypassing Issuer solver selectors. This is fixed by removing create for challenges.acme.cert-manager.io and create/patch/update for orders.acme.cert-manager.io from that role. Upgrade to v1.19.6 to close this HIGH severity RBAC gap (GHSA-8rvj-mm4h-c258).

• securityGo toolchain updated to 1.25.11 for three CVEs

  Applies unconditionally in v1.19.6: Go is updated to 1.25.11, fixing CVE-2026-27145, CVE-2026-42504, and CVE-2026-42507. No action needed beyond upgrading.

• breakingcert-manager-edit ClusterRole loses Challenge/Order create permissions

  If any tooling or workflow creates Challenge or Order resources directly, outside the normal Certificate to CertificateRequest to Order to Challenge flow, it will lose that access after upgrading and needs those permissions granted explicitly.

v1.20.3 is a security release primarily addressing a HIGH-severity RBAC over-permission (GHSA-8rvj-mm4h-c258) in the cert-manager-edit ClusterRole, and updating Go to v1.26.4 for three CVEs. Operators should audit any workflows that create Challenge or Order resources directly before upgrading, as those permissions are now removed.

• securityRBAC over-permission in cert-manager-edit ClusterRole (GHSA-8rvj-mm4h-c258)

  GHSA-8rvj-mm4h-c258 (HIGH): the cert-manager-edit aggregate ClusterRole previously allowed namespace users to create Challenge and Order resources directly, bypassing Issuer solver selectors. Fixed in v1.20.3. Upgrade immediately if untrusted namespace users exist in your cluster.

• securityGo runtime updated to v1.26.4 (3 CVEs)

  Go updated to v1.26.4, patching CVE-2026-27145, CVE-2026-42504, and CVE-2026-42507. No additional configuration is required; upgrading cert-manager picks up the fix.

• breakingBreaking RBAC change: direct Challenge and Order creation removed from cert-manager-edit

  cert-manager-edit no longer grants create on challenges.acme.cert-manager.io, or create/patch/update on orders.acme.cert-manager.io. If any tooling or CI workflow creates Challenge or Order objects directly (outside the normal Certificate → CertificateRequest → Order → Challenge flow), those calls will fail after upgrading. Review and update any such automation before rolling out v1.20.3.

Fluentd v1.19.3 is a security and hardening patch that tightens input validation in out_http, buffer/in_http, and output tag handling, and locks down default access for in_monitor_agent and in_debug_agent. The rest of the release is routine bug fixes and small dependency/CI cleanup.

• securityStricter input validation across out_http, buffer/in_http, and output tag paths

  Applies unconditionally in v1.19.3. out_http now strictly validates hostnames for dynamic endpoints, buffer and in_http enforce size limits on decompressed payloads, and output enforces strict path boundary checks on tag values. Together these close off request-smuggling and decompression-bomb style abuse paths; review any configs that build hosts, tags, or payload sizes from untrusted input.

• breakingin_monitor_agent and in_debug_agent defaults now expose less by default

  Applies unconditionally in v1.19.3. in_monitor_agent changes default visibility of config, retry, and debug info, and in_debug_agent now accepts connections only from the local machine by default. If you relied on remote access to either agent or exposed their full output externally, update firewall rules or explicit config options to restore the access you need.

Dragonfly v2.5.0 is a feature release adding Hugging Face/ModelScope model downloads, a P2P injector webhook, download blocklisting, and broad rate limiting, alongside a breaking removal of deprecated V1 preheat API endpoints and health check consolidation to /healthy. It also ships numerous bug fixes and Nydus subproject improvements.

• breakingV1 preheat API endpoints removed

  Applies unconditionally starting v2.5.0: the deprecated V1 preheat API endpoints have been removed. Any automation or integration still calling V1 preheat routes must migrate to the current preheat API before upgrading.

• breakingHealth checks consolidated to /healthy

  Applies unconditionally starting v2.5.0: health checks are consolidated onto a single /healthy endpoint. Update load balancer, Kubernetes probe, and monitoring configs that point at old or multiple health check paths.

• enhancementEmergency download blocklist

  Available now via Manager console: operators can configure a blocklist to disable specific downloads as an emergency brake. Blocked gRPC calls return PermissionDenied and blocked HTTP proxy calls return FORBIDDEN, useful for incident response.

• enhancementRate limiting across gRPC and client paths

  Available now across Manager/Scheduler gRPC servers and clients: rate limiting covers bandwidth, upload/download requests, and adaptive limiting. Review limits before rollout to avoid unexpected throttling of legitimate traffic.

Istio 1.29.5 is a security release fixing 14 Envoy CVEs under ISTIO-SECURITY-2026-005, with several rated HIGH (DoS, crash, use-after-free). It also repairs a gateway revision-label traffic outage, an ambient ipset health-probe gap, a krt memory leak, and a multicluster controller panic.

• security14 Envoy CVEs fixed (max severity HIGH) — ISTIO-SECURITY-2026-005

  All 14 Envoy CVEs under ISTIO-SECURITY-2026-005 are fixed in 1.29.5. The highest-severity issues include an HTTP/3 QPACK decoding DoS (GHSA-p7C7-7c47-pwch), a compressed-payload inflation bypass of MaxInflateRatio (CVE-2026-48044), a %REQUESTED_SERVER_NAME% formatter crash (CVE-2026-47220), a QUIC content-length validation issue guarded by envoy.reloadable_features.quic_validate_headers_only_content_length (CVE-2026-48743), and a PROXY protocol TLV length mismatch (CVE-2026-47692). Upgrade all 1.29.x deployments to 1.29.5.

• securityOAuth2 filter padding oracle in AES-256-CBC cookie decryption (CVE-2026-47775)

  Operators using the OAuth2 filter with AES-256-CBC cookie encryption should upgrade immediately. The filter now supports AES-256-GCM with a gcm. algorithm marker; plan migration away from AES-256-CBC after upgrading.

• securityext_authz use-after-free with per-route service overrides (CVE-2026-47205)

  Clusters using ext_authz with per-route service overrides are exposed to a use-after-free crash when a downstream connection resets during an in-flight authorization check (CVE-2026-47205). Upgrade to 1.29.5 to eliminate the crash window.

• breakingJSON nesting depth limit reduced to 1000 (CVE-2026-48042)

  JSON parser nesting depth is now capped at 1000 by default (down from 10000). If any Envoy configuration relies on deeper nesting, the old limit can be restored temporarily by setting the runtime flag envoy.reloadable_features.limit_json_parser_nesting_depth to false, but plan to eliminate deep nesting.

Istio 1.30.2 is a security-plus-bugfix patch release, its main content is fixes for 14 Envoy CVEs (one high-severity HTTP/3 DoS via QPACK, others medium to high) under ISTIO-SECURITY-2026-005, alongside operator-relevant changes to pilot-agent metric merging and new AuthorizationPolicy trust-domain fields.

• security14 Envoy CVEs fixed under ISTIO-SECURITY-2026-005, several high severity

  Applies to all 1.30.x deployments. Fixed a DoS in the HTTP/3 stack via QPACK blocked decoding, plus a JSON parser change limiting nesting depth to 1000 (previously the 10K fallback is reachable only if envoy.reloadable_features.limit_json_parser_nesting_depth is set to false) and a headers-only content-length validation fix gated by envoy.reloadable_features.quic_validate_headers_only_content_length. Upgrade to 1.30.2.

• securityMultiple use-after-free and crash fixes across ext_authz, ext_proc, gRPC stats, and OAuth2 filters

  Applies when ext_authz per-route service overrides are active and a downstream connection resets mid-authorization-check (CVE-2026-47205), or when ext_proc/gRPC stats/OAuth2 filters are in use. Upgrade to close use-after-free and padding-oracle issues; the OAuth2 filter now also supports AES-256-GCM cookie encryption via a gcm. marker.

• breakingpilot-agent metric merging now rejects protobuf content type

  Applies to deployments using pilot-agent metric merging (PILOT_AGENT_MERGE_ENVOY_STATS). Allowed content types are now restricted to text/plain and application/openmetrics-text; protobuf-based scraping of merged metrics will break. Check scrape configs before upgrading.

• enhancementAuthorizationPolicy gains trust-domain matching

  Applies to AuthorizationPolicy authors who want mTLS trust-domain-based access control. New trustDomains and notTrustDomains fields on Source let policies match or exclude requests by the trust domain in the peer certificate.

Envoy v1.38.3 is a security-only release patching 15 CVEs plus one upstream wasmtime CVE. Two issues are rated critical: a TLS SAN NUL-byte auth bypass (CVE-2026-47778) and HTTP/3-to-HTTP/1 request smuggling (CVE-2026-48743); the remaining 13 are rated high, covering crashes, a zip-bomb DoS, a PKCE padding oracle, a 65 KB upstream data spillover, and a heap buffer overflow.

• securityCRITICAL: TLS SAN NUL-byte auth bypass (CVE-2026-47778)

  CVE-2026-47778 (GHSA-f8x4-rw5x-f3r7): a NUL byte embedded in a TLS SAN causes truncation during comparison, allowing an attacker to bypass certificate-based authentication. Upgrade to v1.38.3 unconditionally.

• securityCRITICAL: HTTP/3 to HTTP/1 request smuggling (CVE-2026-48743)

  CVE-2026-48743 (GHSA-8phg-2h2q-jgxf): a headers-only HTTP/3 request carrying a nonzero Content-Length can be smuggled into a downstream HTTP/1 backend. Any deployment proxying HTTP/3 to HTTP/1 is affected. Upgrade to v1.38.3.

• securityHIGH: PROXY Protocol v2 upstream data spillover (CVE-2026-47692)

  CVE-2026-47692 (GHSA-wh36-hm39-mm3r): the PROXY Protocol v2 header generator leaks "skipped" TLVs, spilling up to 65 KB of attacker-controlled data into the upstream application stream. Upgrade to v1.38.3.

• breakingIntel DLB connection balancer extension removed

  The contrib extension envoy.network.connection_balance.dlb (Intel DLB connection balancer) has been disabled at the Bazel build layer in v1.38.3 due to a broken source archive. Any deployment that relied on this extension must migrate to an alternative connection balancer.

• breakingTLS certificate Brotli compression now disabled by default

  The runtime guard envoy.reloadable_features.tls_certificate_compression_brotli is now disabled by default. When off, QUIC uses zlib-only certificate compression and TCP TLS performs no certificate compression. If you depended on Brotli compression for TLS certificates, re-enable this flag explicitly after upgrading.

Envoy v1.37.5 is a security-only release fixing 15 Envoy CVEs/GHSAs and one upstream dependency CVE, with severities reaching critical. Three critical issues stand out: HTTP/3-to-HTTP/1 request smuggling, TLS SAN auth bypass via embedded NUL bytes, and a PROXY Protocol v2 spillover that puts attacker-controlled data into upstream streams. All v1.37.x deployments should upgrade without delay.

• securityCritical: HTTP/3 to HTTP/1 request smuggling (CVE-2026-48743)

  CVE-2026-48743 (GHSA-8phg-2h2q-jgxf): HTTP/3 requests with headers-only and a nonzero Content-Length can smuggle requests into HTTP/1 upstream connections. Applies to any deployment terminating HTTP/3 and proxying to HTTP/1 backends. Upgrade immediately.

• securityCritical: TLS SAN NUL-byte auth bypass (CVE-2026-47778)

  CVE-2026-47778 (GHSA-f8x4-rw5x-f3r7): A TLS SAN containing an embedded NUL byte is truncated at that byte during comparison, allowing a certificate to match patterns it should not. Any deployment making auth decisions based on TLS SAN validation is affected. Upgrade immediately.

• securityCritical: PROXY Protocol v2 TLV spillover into upstream stream (CVE-2026-47692)

  CVE-2026-47692 (GHSA-wh36-hm39-mm3r): The PROXY Protocol v2 header generator emits TLVs that were supposed to be skipped, spilling up to 65 KB of attacker-controlled data into the upstream application stream. Affects deployments that generate PROXY Protocol v2 headers toward upstreams. Upgrade immediately.

• securityHigh: Two OAuth2 filter vulnerabilities (padding oracle + UAF)

  CVE-2026-47775 and CVE-2026-48090 (GHSA-396h-jpq4-vc7p, GHSA-3cj2-c63f-q26f): The OAuth2 filter has a PKCE code-verifier padding oracle and a use-after-free risk when an async token fetch completes after stream teardown. Both affect deployments using envoy.filters.http.oauth2.

• securityHigh: HTTP/3 QPACK blocked-decoding DoS (GHSA-p7c7-7c47-pwch)

  HTTP/3 QPACK blocked decoding (GHSA-p7c7-7c47-pwch) can be exploited to cause a DoS against the HTTP/3 stack. Affects any deployment serving HTTP/3.

• securityHigh: grpc_stats filter segfault on Connect protocol (CVE-2026-47204)

  CVE-2026-47204 (GHSA-3jxh-8p6x-7pf6): The grpc_stats filter segfaults on Connect-protocol requests routed to direct_response routes. Affects deployments using envoy.filters.http.grpc_stats.

• breakingIntel DLB connection balancer contrib extension disabled at build time

  The envoy.network.connection_balance.dlb contrib extension (Intel DLB connection balancer) has been disabled at the Bazel build layer across all platforms due to a source archive breakage. Deployments relying on this extension must pursue a local workaround or alternative balancing strategy until the source issue is resolved.

Envoy v1.36.9 is a security-only release patching 15 CVEs (2 rated critical, the rest high or medium) across TLS SAN handling, HTTP/3, PROXY protocol, JSON parsing, OAuth2, ext_proc, grpc_stats, zstd, DNS, and statsd. Operators should upgrade immediately, as two of the vulnerabilities -- TLS SAN auth bypass and HTTP/3 request smuggling -- are unconditional or broadly exploitable.

• securityCRITICAL: TLS SAN NUL-byte auth bypass (CVE-2026-47778 / GHSA-f8x4-rw5x-f3r7)

  A TLS SAN containing an embedded NUL byte causes the SAN to be truncated, bypassing certificate-based authentication checks. All deployments are affected. Upgrade to v1.36.9 immediately.

• securityCRITICAL: HTTP/3-to-HTTP/1 request smuggling (CVE-2026-48743 / GHSA-8phg-2h2q-jgxf)

  A headers-only HTTP/3 request with a nonzero Content-Length can be smuggled into HTTP/1 backends, enabling request smuggling attacks. Affects any deployment proxying HTTP/3 to HTTP/1 backends.

• securityHIGH: PROXY Protocol v2 upstream data spillover (CVE-2026-47692 / GHSA-wh36-hm39-mm3r)

  PROXY Protocol v2 header generation leaks up to 65 KB of attacker-controlled data into the upstream application stream via improperly emitted TLVs. Affects deployments using the PROXY Protocol v2 header generator.

• securityHIGH: Stack overflow on deeply nested JSON (CVE-2026-48042 / GHSA-f24p-rxw2-g6pv)

  Deeply nested JSON input triggers a stack overflow in the object destructor, crashing the proxy. All deployments are affected.

• securityHIGH: wasmtime dependency CVE (CVE-2026-47261)

  The wasmtime dependency was bumped to resolve CVE-2026-47261. All deployments using Wasm extensions are affected.

• breakingDLB connection balancer extension disabled (capability removed)

  The contrib Intel DLB connection balancer extension (envoy.network.connection_balance.dlb) is now disabled at the Bazel layer for all platforms due to a source archive breakage. Builds that previously enabled this extension will lose it silently.

Envoy v1.35.13 is a major security release fixing 13 CVEs, two of them critical: a PROXY protocol v2 TLV spillover leaking up to 65KB into the upstream stream, and a TLS SAN NUL-truncation bug that can bypass certificate-based auth. It also disables the contrib Intel DLB connection balancer extension due to a build breakage.

• securityPROXY Protocol v2 TLV spillover (CVE-2026-47692)

  Applies to all v1.35.13 deployments running PROXY protocol v2 header generation upstream. CVE-2026-47692 lets an attacker's "skipped" TLVs leak up to 65KB into the upstream application stream. Upgrade immediately if you use this feature.

• securityTLS SAN NUL truncation enables auth bypass (CVE-2026-47778)

  Applies to deployments validating client certificates by SAN. CVE-2026-47778 lets an embedded NUL byte in a TLS SAN truncate the parsed name, allowing certificates to match unintended hostnames and bypass authentication. Upgrade immediately.

• security11 more high-severity CVEs across core filters and HTTP/3

  Eleven additional high-severity CVEs are fixed across ext_proc, router internal redirects, OAuth2 (two separate issues: padding oracle and a use-after-free from late async token completion), zstd decompression, grpc_stats, JSON parsing, the DNS filter, HTTP/3 (content-length validation and a separate QPACK-based DoS), and TcpStatsdSync. Each is a crash, DoS, or memory-safety bug tied to a specific filter or codec; upgrade to pick up all fixes, or check which filters you run to prioritize.

• securityWasmtime dependency bump fixes CVE-2026-47261

  A separate wasmtime CVE (CVE-2026-47261) is fixed by bumping the com_github_wasmtime dependency. Applies to deployments using the Wasm filter.

• breakingIntel DLB connection balancer extension disabled

  The contrib extension envoy.network.connection_balance.dlb (Intel DLB connection balancer) is disabled at the Bazel build layer for all builds and platforms due to a source archive breakage. Anyone building Envoy with this extension enabled needs an alternate connection balancer.

v0.155.0 is a breaking-change release for operators: seven stabilized feature gates are removed, memory_limiter metrics get a new prefix, and two config/service APIs are deprecated in favor of snapshot-based replacements. The rest is tooling work on mdatagen and the newly relocated schemagen CLI.

• breakingSeven stabilized feature gates removed

  Applies to any collector build referencing confighttp.framedSnappy, configoptional.AddEnabledField, confmap.newExpandedValueSanitizer, exporter.PersistRequestContext, otelcol.printInitialConfig, telemetry.UseLocalHostAsDefaultMetricsAddress, or pdata.enableRefCounting via --feature-gates flags or config. These gates are gone in v0.155.0; startup will fail if they're still referenced. Drop them from your CLI args and config before upgrading.

• breakingmemory_limiter metrics renamed with processor prefix

  Applies to anyone scraping or alerting on otelcol_processor_* metrics from the memory_limiter processor. They're renamed to otelcol_processor_memory_limiter_* to disambiguate from other processors. Update dashboards, alert rules, and metric queries to the new names.

• breakingmdatagen reaggregation_enabled setting removed

  Applies to custom components using mdatagen metadata.yaml with reaggregation_enabled. The setting is removed; per-metric reaggregation config is now always generated. Old files with the field are still accepted but the value is ignored, so check generated output matches expectations.

• breakingConfig watcher APIs deprecated for snapshot-based equivalents

  Applies to extension or core code using service.Settings.CollectorConf or extensioncapabilities.ConfigWatcher. Both are deprecated in favor of service.Settings.ConfigSnapshot and extensioncapabilities.ConfigSnapshotWatcher. No forced migration yet, but plan to move to the snapshot-based APIs.

Crossplane v2.3.3 patches a TOCTOU security flaw in OCI package signature verification (GHSA-mf7q-r4rv-jv94) and fixes a namespace injection bug in `crossplane render`. Routine CVE bumps for Go deps included.

• securityUpgrade immediately to fix OCI signature TOCTOU (GHSA-mf7q-r4rv-jv94)

  The TOCTOU flaw means a compromised or malicious OCI registry could serve unsigned package content after passing signature verification. Any Crossplane deployment that installs packages from OCI registries is exposed. Upgrade to v2.3.3 — this fix lives in crossplane-runtime and is bundled in this release. Review the crossplane-runtime v2.3.3 advisory for full technical details and assess whether any packages installed on affected versions should be reinstalled.

• securitygolang.org/x/net and x/sys CVE patches — rebuild or upgrade

  The apis module now pulls updated golang.org/x/net and golang.org/x/sys. If you build Crossplane from source or vendor these deps in your own providers/functions, update your dependency pins to pick up the same CVE fixes. Pre-built images in v2.3.3 already include the patched versions.

• breakingcrossplane render namespace behavior changed for namespaced XRs

  If you use `crossplane render` in CI pipelines or local testing with namespaced XRs, the injected resource refs no longer carry a namespace. This matches real reconciler behavior and fixes breakage with strict-schema functions (e.g., KCL-generated bindings). Re-run your render tests after upgrading to confirm output changes don't mask real issues in your composition logic.

OpenKruise v1.9.0 adds several useful workload features and fixes a pile of stability bugs — including CPU-spike and panic fixes that affect anyone running ImagePullJobs or SidecarSets in production.

• breakingString-formatted maxUnavailable/maxSurge values now rejected by webhook

  Values like maxUnavailable: '5' (a quoted integer string) are now rejected at admission. If your manifests or Helm charts set these as strings instead of integers, the webhook will block them after upgrade. Audit your CloneSet and related workload specs before upgrading to v1.9.0.

• enhancementFix ImagePullJob high CPU before upgrading large clusters

  A cascading reconcile amplification bug was causing ImagePullJob controllers to spike CPU. If you've been seeing unexpected controller-manager CPU pressure, this is likely the cause. Upgrading to v1.9.0 should resolve it — no config changes needed.

• enhancementCloneSet progressDeadlineSeconds and OnDelete strategy now available

  CloneSet now supports progressDeadlineSeconds to surface stalled rollouts, and an OnDelete update strategy for manual upgrade control. Teams doing canary or phased rollouts with CloneSet should evaluate both options — progressDeadlineSeconds in particular fills a gap that previously required external monitoring to detect stuck updates.

Lima v2.1.3 is a security-focused patch release fixing two CVEs: a privilege escalation in QEMU VMs and multiple containerd CVEs bundled via nerdctl v2.3.3. Upgrade promptly.

• securityPatch QEMU privilege escalation (CVE-2026-53657) immediately

  Any unprivileged user inside a QEMU-backed Lima VM could gain root inside that VM by abusing the guest agent socket. If you share Lima VMs among multiple users or run untrusted workloads, this is high priority. Upgrade to v2.1.3 and restart affected VMs.

• securitycontainerd CVE batch via nerdctl v2.3.3 — upgrade required

  Five CVEs in containerd are fixed in the bundled nerdctl v2.3.3 (containerd v2.3.2). If your Lima instances run container workloads, the old containerd is exposed. Upgrade Lima and recreate or restart VMs to pick up the new nerdctl binary.

• breakingcontainerd.user defaults changed for non-Linux guests

  Lima no longer sets containerd.user=true by default on non-Linux guests (macOS, Windows). If you relied on rootless containerd on those platforms without explicitly setting this, validate your VM config after upgrading and set containerd.user=true explicitly if needed.

Argo CD 3.4.4 is a patch release focused on stability fixes for health checks, RBAC, diffs, and template rendering. Update if you run multi-namespace setups or use Dex authentication.

• breakingRBAC regression fix for multi-namespace setups

  A prior release introduced a regression in RBAC enforcement for project-scoped resources in multi-namespace architectures. This patch restores correct behavior. If you use project-level RBAC restrictions across multiple namespaces, verify that access controls work as expected after upgrading, particularly around project-scoped resource visibility.

• enhancementHealth check reliability improvement

  The PromotionStrategy health check was incorrectly reporting Progressing after no-op re-hydration, blocking deployments. This is now fixed. If you use PromotionStrategy with ArgoCD, upgrade to avoid stalled health states that require manual intervention to clear.

• enhancementServer-side diff regression resolved

  A recent change broke server-side diffs on new objects, causing errors. This patch restores diff functionality. If you rely on server-side diffs in your deployment pipeline (especially for new resources), apply this update to restore normal diff behavior.

Argo CD v3.3.12 is a maintenance release fixing sync state tracking, cluster observer locking, configuration parsing, and git repository depth handling—no breaking changes.

• securityDex authentication now handles special characters in passwords

  Dex password parsing previously failed if passwords contained dollar signs, potentially locking users out. If you use Dex for OIDC and recently added or updated passwords with special characters, upgrade to v3.3.12 immediately to restore access. Test Dex login after updating.

• enhancementFix PromotionStrategy stuck state if using Argo Rollouts

  If you run Argo CD with Argo Rollouts and use PromotionStrategy, applications may have remained stuck in Progressing after configuration reloads even when nothing changed. Update to v3.3.12 to resolve this. Check your application health status post-upgrade to confirm recovery.

Helm v4.2.2 reverts a premature exit fix in WaitForDelete that was causing test failures. A one-line patch addressing a race condition in status observation.

• breakingWaitForDelete race condition re-introduced—test suite hangs may return

  The patch that fixed a race condition in WaitForDelete (where the status observer canceled the watch too early) has been reverted in v4.2.2. If you run comprehensive Helm test suites or rely on WaitForDelete behavior, you may see intermittent hangs or timeouts again. Monitor test runs for flakiness after upgrading. A proper fix is likely incoming in a future patch—do not stay on v4.2.2 longer than necessary.

Strimzi 1.0.1 patches two CVEs and disables cross-namespace Entity Operator watching by default. The v1-only CRD API requirement from 1.0.0 remains enforced — no v1beta2/v1beta1/v1alpha1.

• securityPatch two CVEs — upgrade promptly

  CVE-2026-55225 and CVE-2026-55226 are fixed in 1.0.1. Details are in the Strimzi security advisories. Given both are fixed in the same patch release, treat this as a security-motivated upgrade and prioritize it over waiting for a convenient maintenance window.

• breakingEnable cross-namespace watching explicitly if you use watchedNamespace

  If your Kafka CR's Entity Operator section has watchedNamespace set to a namespace different from the Kafka cluster's namespace, the Topic or User Operator will stop watching it after upgrading. Before or immediately after upgrading, set STRIMZI_ENTITY_OPERATOR_WATCHED_NAMESPACE_ENABLED=true in your Cluster Operator deployment. Check all Kafka CRs across namespaces — missing this will silently break topic/user reconciliation.

• breakingComplete CRD migration to v1 before upgrading

  Strimzi 1.0.1 (like 1.0.0) only accepts v1 CRDs. If you skipped the migration step when moving to 1.0.0, or if you are upgrading from an older release directly, run the CRD conversion procedure documented in the Strimzi 1.0.1 deploying guide before applying the new operator manifests. Applying the operator without converted CRDs will cause reconciliation failures.

v1.18.0 patches two auth security bugs and fixes MySQL case-sensitivity. MySQL users must plan a maintenance window before upgrading — auto-migration will lock tuple/changelog tables.

• securitySilent OIDC audience bypass — check your auth config before upgrading

  If authn.oidc.audience was previously omitted, any validly-signed token from your trusted issuer was accepted regardless of intended audience. After upgrading, OpenFGA will refuse to start without both issuer and audience set. Verify your deployment config has authn.oidc.audience explicitly set, or the service will not come up.

• securityPreshared key timing side-channel fixed — no action needed if you rotate keys

  The prior map lookup for preshared key auth could reveal valid key bytes via timing differences. Now fixed with constant-time comparison. No immediate action required, but if this key has been exposed to untrusted network paths for a long time, rotating the preshared key is a reasonable precaution.

• breakingMySQL users: do not auto-migrate on startup

  Migration 008 acquires a shared lock on tuple and changelog tables. On large datasets this can block Write operations for an extended period. Read the operator runbook at the collation_migrations.md guide, schedule a maintenance window, and run the migration manually rather than letting OpenFGA auto-migrate on startup.

Rook v1.20.1 is a patch release focused on Ceph operator stability: Ceph CSI driver compatibility, OSD major-version handling, and CSI-addons disabled by default to reduce unexpected behavior.

• breakingCSI-addons disabled by default — check cluster dependencies

  CSI-addons (snapshot, clone, and expansion features) are now disabled by default in v1.20.1. If your workloads rely on these features, you must explicitly enable them in the Ceph cluster spec or your RWX volume snapshots and clones will fail. Review helm values and operator configs to enable addons before upgrade if needed.

• enhancementUse node labels for OSD device class assignment instead of manual mapping

  Instead of hardcoding device class assignment per node, you can now use Kubernetes node labels (like `disk-type: ssd`) to automatically classify OSDs. This simplifies scaling clusters with mixed hardware. Apply node labels, then reference them in the Ceph cluster spec to avoid per-node configuration drift.

• enhancementOSD require-osd-release is now enforced after major Ceph upgrades

  After a major Ceph version upgrade (e.g., Quincy to Reef), Rook now sets `require-osd-release` automatically to prevent old OSDs from rejoining before the cluster is ready. Before upgrade, ensure all OSDs can be updated together; if you have pinned some OSDs to an older version, they will be blocked from joining until unpinned.

v1.52.0 has three breaking changes (discovery API default, immediate stitching removed, BUI union types) alongside significant catalog PostgreSQL performance fixes and a batch of UX and reliability improvements.

• breakingCheck discovery.endpoints and remove immediate stitching config

  Two breaking changes need attention before upgrading. First, if you use `discovery.endpoints` with internal-only string targets, convert them to object form with `target.internal` set to the internal URL and `target.external` set to a browser-reachable URL — otherwise the frontend will try to reach internal addresses directly. Second, remove `catalog.stitchingStrategy.mode: 'immediate'` from your config; it's now a no-op and will log deprecation warnings. Both changes are straightforward config edits.

• breakingMigrate ComboboxProps and SelectProps extensions before upgrading

  `ComboboxProps` and `SelectProps` are now union types. If your codebase has interfaces that extend either type, they will fail to compile — switch them to type intersections. Also audit CSS selectors that target list content as a direct child of `.bui-SelectPopover`, as that structure changed. These are frontend component-layer changes with no runtime fallback.

• enhancementCatalog query performance improvements for PostgreSQL — no action needed, just upgrade

  Several PostgreSQL-specific query planner fixes ship in this release: multi-column statistics on the search table, a dropped redundant index, tuned vacuum thresholds, and a split count/list query. If your catalog list views are slow with large entity counts, this upgrade is worth prioritizing. No config changes needed — the improvements apply automatically via migration. Teams on MySQL/SQLite won't see the same gains.

KubeVirt v1.8.4 is a patch release fixing a gRPC connection leak in virt-handler that caused memory growth, patching CVE-2026-35469 in spdystream, and adding missing metrics/alerts.

• securityPatch CVE-2026-35469 by upgrading to v1.8.4

  The moby/spdystream dependency carried CVE-2026-35469 (GHSA-pc3f-x583-g7j2). If you're on any v1.8.x release before v1.8.4, upgrade now. Check your internal scanner results for this CVE to confirm exposure before and after the upgrade.

• breakinggRPC connection leak fix may change virt-handler resource footprint

  The connection leak in GetLauncherClient caused unbounded memory and goroutine growth when multiple controllers raced on the same VMI. After upgrading, virt-handler memory usage should drop noticeably in clusters with high VMI churn. If you have memory-based alerts or resource limits tuned to the leaked baseline, revisit those thresholds post-upgrade.

• enhancementNew metrics and alerts for virt components — update dashboards

  Missing metrics, recording rules, and alerts were added for virt components. Review what's new against your existing Prometheus/Alertmanager setup and add any new alerts to your runbooks. This is a good time to audit alert coverage gaps you may have been living with.

Cilium v1.19.5 is a bugfix-heavy patch release addressing policy enforcement bugs, Gateway API panics, and load-balancer correctness issues. No CVEs, but several fixes warrant attention before upgrading.

• breakingMigrate away from `l2podAnnouncements.interface` Helm value before upgrading

  The `l2podAnnouncements.interface` Helm key was removed because it wrote a configmap entry the agent no longer reads, causing crash-loops when L2 pod announcements are enabled. Before upgrading, check your Helm values for this key and replace it with `l2podAnnouncements.interfacePattern`. Clusters using the old key will fail to start after the upgrade.

• breakingCheck CiliumEgressGatewayPolicy if you use NamespaceSelector

  A bug caused the NamespaceSelector field in CiliumEgressGatewayPolicy to be silently corrupted, making those rules ineffective. After upgrading to 1.19.5, verify that egress policies with NamespaceSelector are actually being enforced — traffic that was previously leaking may now be blocked.

• enhancementVerify ipBlock network policies if you rely on selectorless rules

  A bypass was fixed where wildcard namespace selectors could skip ipBlock rules without a pod selector. If you have CiliumNetworkPolicies using ipBlock without explicit namespace selectors, review them post-upgrade to confirm enforcement behavior matches your intent — the fix may change observed traffic patterns.

Cilium 1.18.11 is a patch release focused on stability, fixing memory leaks, socket handling crashes, and Gateway API routing bugs. No breaking changes; safe to adopt for teams running 1.18.x in production.

• securityApply patch if you see agent crashes in socket handling

  Cilium 1.18.11 fixes a nil pointer dereference in filterAndDestroySockets that could crash the agent. If you have seen sudden agent pod restarts or panic logs mentioning socket filtering, upgrade promptly. This is a stability issue that affects 1.18.10 and earlier.

• breakingVerify MTU settings on non-Cilium interfaces after upgrade

  1.18.11 changes MTU handling to skip interfaces not managed by Cilium. If your infrastructure depends on Cilium adjusting MTU on external interfaces (e.g., host-managed veth pairs), test this upgrade in a staging environment first. Mark interfaces with altname to ensure they are recognized as Cilium-owned; otherwise, MTU will not be modified.

• enhancementAdopt if running Kubernetes Gateway API with TLS routes

  This patch fixes two distinct bugs in Gateway API support: weighted backend routing for TLSRoute passthrough and mixed listener handling. If you use Gateway API with TLS termination or passthrough, this upgrade eliminates silent routing failures. No configuration changes needed; test existing Gateway API resources in a staging cluster first.

Dapr v1.18.1 is a focused bug-fix release for the workflow engine, patching five issues including reminder leaks, permanent sidecar unavailability after config reload, and stuck workflows.

• breakingHelm users: workflow concurrency limits were silently ignored before this release

  If you set globalMaxConcurrentWorkflowInvocations, globalMaxConcurrentActivityInvocations, or the per-name limit fields on a Configuration resource and installed via Helm, those limits were never enforced. After upgrading to 1.18.1 the CRD schema is correct and limits will actually take effect. Review your configured values before upgrading in production — limits that were previously ignored will now be applied, which could throttle throughput if the values are too conservative.

• enhancementUpgrade if you run agentic or long-running loop workflows

  Three bugs in this release specifically affect workflows that loop via ContinueAsNew or repeatedly await the same external event name — patterns common in agentic workloads. The reminder leak accumulates garbage reminders over a workflow's lifetime; the ContinueAsNew deadlock causes indefinite hangs with no timeout or error. Both are fixed here. If you see workflows stuck in RUNNING or observe spurious wake-ups in loop workflows, upgrade to 1.18.1.

• enhancementConfig hot-reload now works reliably for workflow sidecars

  Before this fix, any SIGHUP-triggered config reload on a sidecar hosting workflow workers could permanently break the sidecar — port 50001 stops accepting connections and the only recovery was a pod restart. This is now fixed. If your team uses config hot-reload in any workflow-hosting deployment, 1.18.1 should be treated as a required upgrade.

Linkerd edge-26.6.2 patches policy and profile validation bugs, tightens namespace restrictions on external workloads, and upgrades to Envoy proxy v2.357.0 with unified failure accrual tuning.

• securityProfile validation now rejects nil configs

  The nil check in profile validation closes a path to crashes on malformed configs. No immediate action needed, but if your automation generates profiles programmatically, ensure they always have valid structure before applying them.

• breakingNamespace-restricted external workloads may affect routing

  External workloads and endpoints now honor namespace boundaries. If you cross namespaces with external workloads in your Linkerd setup, verify routing still works after upgrade. Update any policies that rely on previous (unrestricted) cross-namespace access.

• enhancementTune load balancing with unified failure accrual

  Unified failure accrual and response-penalty load biasing gives you finer control over how proxies handle slow or failing endpoints. Test this in a canary namespace first—adjust failure thresholds and penalty weights to match your SLOs. Check Linkerd docs for new policy knobs.

Dapr 1.17.10 fixes a bug where resiliency retry policies with `matching` rules on pubsub publish operations were silently ignored, causing terminal errors to be retried unnecessarily. Teams using pubsub with resiliency retry configuration should upgrade to enforce intended retry behavior.

• breakingRetry behavior change for pubsub publish errors

  If you have resiliency policies with `matching` rules targeting pubsub outbound retry, publish errors will now be classified correctly and non-retriable errors will stop retrying immediately instead of exhausting maxRetries. Verify your resiliency configuration expectations: terminal errors (e.g., 401 Unauthorized, 404 Not Found) will no longer consume retry attempts. If your application relied on the old behavior of always retrying publish, adjust your retry policy or remove the `matching` constraint.

• enhancementEnforce resiliency policy intent for pubsub

  Upgrade to 1.17.10 if you've configured resiliency policies with explicit `matching` codes for pubsub publish — your configuration will now work as intended. This brings pubsub publish in line with service invocation, output bindings, and other operations that already respect retry `matching`. No code changes needed; the fix is transparent.

Dapr 1.16.16 fixes two downgrade-specific bugs: Sentry certificate signing failures after a 1.18→1.16 rollback, and Helm StatefulSet storage conflicts when downgrading from 1.17/1.18.

• breakingFollow specific Helm flags when downgrading to 1.16.16

  Use --reset-values and explicitly pass the original install values. Do NOT use --reuse-values — it carries over 1.17/1.18 chart defaults that are invalid in 1.16, including placement disseminateTimeout=8s (1.16 only accepts 1s–3s), which will cause placement to fail at startup. If you skipped deleting the scheduler StatefulSet before downgrading, you may still need to pass --set dapr_scheduler.cluster.storageSize=<current size> or delete the StatefulSet with --cascade=orphan.

• breakingClusters downgraded from 1.18 to 1.16 need this patch to fix Sentry

  If you rolled back from Dapr 1.18 to any earlier 1.16.x release, Sentry is likely crashing on startup due to the Ed25519/ECDSA key type mismatch in the trust bundle. Upgrade to 1.16.16 to resolve it — no manual certificate rotation is required after upgrading.

KServe v0.19.0 is a large release focused heavily on LLMInferenceService (LLMISvc) maturity: better observability, autoscaling, routing, and a migration path for llm-d v0.6 upgrades. Two CVEs are patched.

• securityPatch two CVEs before upgrading or apply independently

  v0.19.0 patches a vLLM/Pillow CVE and pins azure-core>=1.38.0 for CVE-2026-21226. If you are running KServe with vLLM runtimes or azure-core dependencies, upgrade to v0.19.0 or manually apply the azure-core pin to your environment. Check your current Pillow version in custom server images as well.

• breakingRouter splitter off-by-one fix changes traffic split behavior

  The pickupRoute random range had an off-by-one error. After upgrading, traffic percentages in InferenceService splitter configurations will be computed correctly, which means observed traffic distribution will shift if your weights were tuned around the buggy behavior. Verify split configurations in staging before rolling to production.

• enhancementLLMISvc autoscaling and status visibility are now production-ready

  HPA/KEDA scaling conditions are now visible in LLMISvc status, and readiness transitions emit k8s events. If you are operating LLMInferenceService, add these conditions to your monitoring dashboards and alerting. The --scaling flag in kserve-install.sh simplifies enabling autoscaling during fresh installs.