Hello all, We are excited to announce the release of Apache Ozone 2.1.1!
https://ozone.apache.org/release-notes/2.1.1 Documentation: https://ozone.apache.org/docs/ Docker image: https://hub.docker.com/r/apache/ozone/tags Javadoc: https://javadoc.io/doc/org.apache.ozone Github release page https://github.com/apache/ozone/releases/tag/ozone-2.1.1 Critical bug fixes (upgrade strongly recommended)HDDS-14858 <https://issues.apache.org/jira/browse/HDDS-14858> — OM requests fail on JDK 11 with ClassNotFoundException: java.lang.constant.Constable Impact: If you run Ozone 2.1.0 on JDK 11 or lower, OM operations such as ozone sh key put can fail with: RemoteException: java/lang/constant/ConstableCaused by: java.lang.ClassNotFoundException: java.lang.constant.Constable Who is affected: Anyone on 2.1.0 + JDK ≤ 11. This is a regression from AspectJ bytecode generation referencing JDK 12+ APIs. Action: Upgrade to 2.1.1 if you cannot move to JDK 17+. HDDS-14778 <https://issues.apache.org/jira/browse/HDDS-14778> — ozone-filesystem-shaded protobuf corruption breaks ofs:// clients Impact: With TRACE logging enabled, Ozone filesystem client operations via ofs:// can fail at class load time: ExceptionInInitializerError at OzoneManagerProtocolProtos.Caused by: InvalidProtocolBufferException: Protocol message tag had invalid wire type Root cause: Over-broad Maven shade relocation rules (io, kotlin, info, etc.) corrupted protobuf descriptor bytes inside the shaded JAR. The bug is latent and can reappear after proto changes. Who is affected: Users of ozone-filesystem-shaded (Hadoop/OFS integration), especially with debug/trace logging. Action: Upgrade to 2.1.1 and rebuild/redeploy shaded client JARs. Security / authorization behavior changesHDDS-14898 <https://issues.apache.org/jira/browse/HDDS-14898> & HDDS-14894 <https://issues.apache.org/jira/browse/HDDS-14894> — Missing ACL checks on S3 multipart APIs Impact: Behavior change (tightening). ListParts and ListMultipartUploads previously had no ACL checks. They now enforce authorization like other S3 APIs. Who is affected: - STS users with narrowly scoped tokens (e.g., PutObject-only) that previously could call these APIs - Any workflow that relied on implicit access via multipart upload ownership Action: After upgrade, verify multipart upload workflows and STS session policies. Previously “working” calls may now return 403 Access Denied. HDDS-15064 <https://issues.apache.org/jira/browse/HDDS-15064> — Ranger/STS S3 action–aware authorization Impact: API/authorization model change for STS + Apache Ranger integration. - Adds s3Action to RequestContext so Ranger can restrict permissions by specific S3 action (e.g., distinguish s3:PutObjectTagging from s3:DeleteObjectTagging) - OzoneGrant now carries a Set of allowed S3 actions for inline policies Who is affected: Clusters using STS temporary credentials with Ranger. Authorization becomes more granular; tokens may grant less access than before when S3 actions are explicitly scoped. Action: Coordinate with your Ranger/Ozone plugin team. This was backported specifically so Ranger can consume it upstream in 2.1.1. HDDS-14366 <https://issues.apache.org/jira/browse/HDDS-14366> — Log4j2 bump to 2.25.3 (CVE-2025-68161 <https://github.com/advisories/GHSA-vc5p-v9hr-52mj>) Impact: Fixes TLS hostname verification bypass in Log4j2 Socket Appender (MITM risk for remote log shipping). Who is affected: Only if you use Log4j2 Socket Appender over TLS with hostname verification enabled. Most clusters are unaffected unless they ship logs this way. Notable operational fixes (lower severity)HDDS-14368 <https://issues.apache.org/jira/browse/HDDS-14368> — Recon shows wrong pipelines per container Impact: Recon UI/API incorrectly showed all pipelines for every container instead of the container’s actual WRITE pipeline. Action: Upgrade if you rely on Recon for container/pipeline troubleshooting. HDDS-13069 <https://issues.apache.org/jira/browse/HDDS-13069> — S3 Gateway shutdown error Impact: S3 Gateway logged an IllegalStateException from Weld/Jetty during admin webserver shutdown. Shutdown could appear to fail even though the process was stopping. Action: Cosmetic/operational fix; shutdown now completes cleanly (error is caught and logged).
