// Apache Hadoop extended support
Your Hadoop cluster is older than its support.
So is the ZooKeeper that keeps HDFS highly available.
The Hadoop project has declared every 2.x line before 2.10 end of life, along with 3.0, 3.1 and 3.2. 2.10 and 3.3 have not been formally retired, but 2.10 has had no release since May 2022 and 3.3 none since June 2023. HDFS NameNode HA and YARN ResourceManager HA both run on ZooKeeper, and Hadoop 2.10.2 and 3.2.4 bundle ZooKeeper 3.4.14. OSSeva ships patched, signed Hadoop builds today and patches the ensemble behind them.
Trusted globally by enterprises
Why Hadoop clusters stay where they are
A Hadoop upgrade moves storage, compute and every job that touches them. That is why it waits.
Retired lines get no releases
When the Hadoop community retires a line, issues targeting it are no longer fixed. The 3.2 vote, which passed in December 2023, said so for 3.2.5 in as many words. 3.1 was retired in June 2021 and 2.9 in September 2020.
High availability runs through ZooKeeper
Automatic NameNode failover adds a ZooKeeper quorum and the ZKFailoverController (ZKFC) next to each NameNode. YARN ResourceManager HA uses a ZooKeeper-based elector and, by recommendation, the ZooKeeper state store. Without the ensemble there is no automatic failover.
The bundled ZooKeeper is end of life
Hadoop 2.10.2 and 3.2.4 bundle ZooKeeper 3.4.14, and 3.3.6 bundles 3.6.3. Both ZooKeeper lines are past community end of life, never received a fix for CVE-2023-44981, and ship log4j 1.2.17. Hadoop 3.4 and 3.5 moved to ZooKeeper 3.8.
The dates that matter
2020-03-09
Hadoop 2.8 end of life announced.
2020-09-07
Hadoop 2.9 end of life announced. Every 2.x line before 2.10 is now retired.
2021-06-10
The vote to retire Hadoop 3.1 passes.
2022-05-31
Hadoop 2.10.2 released, still the latest 2.10 release.
2023-06-23
Hadoop 3.3.6 released, still the latest 3.3 release.
2023-12-22
Hadoop 3.2 declared end of life. Its last release, 3.2.4, bundles ZooKeeper 3.4.14.
2026-04-02
Hadoop 3.5.0 released with ZooKeeper 3.8.6.
What OSSeva delivers
Patched HDFS and YARN builds
Backported security fixes for the Hadoop line you run, from 2.x to 3.3, delivered as signed tarballs, Maven artifacts and container images. NameNode, DataNode, ResourceManager and NodeManager are all covered.
The ZooKeeper behind HA
Patched builds of ZooKeeper 3.4, 3.5, 3.6 and 3.7 for the ensemble that ZKFC and ResourceManager HA depend on, available now. Configuration and data stay the same, and members are replaced one at a time with quorum kept.
Upgrade to Hadoop 3.4 or 3.5
A tested path from the patched line to a current release, covering Java versions, client compatibility and failover drills, run as a rolling upgrade where the versions allow it.
Your options, compared
| Option | What you get | Trade-off |
|---|---|---|
| Upgrade to Hadoop 3.4 or 3.5 | Community-maintained releases with a ZooKeeper 3.8 client | Every job, client and connector has to be tested against the new version first. |
| Move to a commercial distribution | Vendor-supported Hadoop | A platform migration and a new subscription. |
| Migrate off Hadoop | A cloud or lakehouse data platform | Data movement and job rewrites, usually over several quarters. |
| OSSeva extended support | Patched Hadoop and ZooKeeper now, and the upgrade later | A subscription while the old line runs. |
| Stay unpatched | Nothing | Advisories stay open on the cluster that holds the most data. |
End-of-life dates from the Hadoop project's votes and announcements on common-dev@hadoop.apache.org and general@hadoop.apache.org. Release dates from hadoop.apache.org. HA design from the HDFS and YARN HA documentation. Bundled ZooKeeper versions from hadoop-project/pom.xml at each release tag.
Frequently asked questions
Which Hadoop versions are end of life?
The project has retired 1.x, 2.0 to 2.9, 3.0, 3.1 and 3.2, the last of them in December 2023. 2.10 and 3.3 have not been formally retired, but their latest releases are 2.10.2 (May 2022) and 3.3.6 (June 2023). 3.4 and 3.5 are the current lines.
What is ZKFC in Hadoop?
The ZKFailoverController. It runs next to each NameNode, checks its health and holds a ZooKeeper lock that decides which NameNode is active. Without a working ensemble, automatic failover stops.
Does Hadoop need ZooKeeper?
For high availability, yes. Manual failover and a single ResourceManager work without it, but almost every production cluster runs HA.
Does OSSeva patch Hadoop inside CDH or HDP?
Yes. CDH and HDP have their own pages, because each distribution ships its own builds of Hadoop and ZooKeeper.
Which support tier fits a Hadoop cluster?
Patch covers signed builds with backported fixes. Assure adds an ensemble audit and an attestation package for auditors. Operate adds 24/7 monitoring and named engineers.
Keep HDFS and YARN patched while the upgrade is planned.
Start with Patch today. Add Assure or Operate when the cluster needs them.