Cloudera Docs

Fixed issues in Cloudera Runtime 7.1.9 SP2 CHF 2

Fixed issues for Cloudera Runtime 7.1.9 SP2 Cumulative Hotfix 2 (CHF 2), released on Aug 31, 2026, provide resolutions for identified bugs across core Cloudera Runtime components.

The following fixes were delivered for CDP Private Cloud Base version 7.1.9.20201-1 (Parcel URL: 7.1.9-1.cdh7.1.9.p20201.82473061):

CDPD-99365: DataNode: Tolerate per-volume health-check latch timeouts before marking volumes failed
Previously, the DataNode health-check process waited on a single latch for all volume checks to complete. If this latch expired due to a transient stall, every pending volume was immediately marked as failed, even if no actual I/O failure occurred. This led to false-positive volume failures.

This issue is now resolved, and a per-volume tolerance for latch timeouts is introduced. A volume is now marked as failed only if it exceeds a configurable number of consecutive timeouts, while genuine I/O failures continue to be reported immediately.

Apache Jira: HDDS-14871

CDPD-101838: Ozone Manager crashes during startup due to a race condition with Ratis snapshot installation

Previously, a race condition occurred when a follower Ozone Manager (OM) daemon started up while concurrently receiving a database snapshot from the Ratis leader. Both the main startup thread and the snapshot installer thread attempted to initialize OzoneDelegationTokenSecretManager simultaneously, which triggered a validation failure. Consequently, the OM daemon terminated unexpectedly during initialization and threw a java.lang.IllegalStateException error in OzoneSecretManager.start() method.

This issue is now resolved. Thread synchronization is now updated to prevent simultaneous initialization, resolving the race condition and allowing the OM daemon to start successfully.

Apache Issue: HDDS-15103
CDPD-99878: Ozone Container Balancer fails to return source DataNodes to the queue after specific move failures
Previously, when the Ozone Container Balancer encountered a move failure due to a container already being moved or unhealthy replication after a move, the source DataNode was not added back to the queue, which led to inefficient balancing operations.

This issue is now resolved, and the source DataNode is returned to the queue in these scenarios, and containers with unhealthy replication are properly excluded from further move attempts.

Apache Issue: HDDS-14906
CDPD-102272: Ozone DataNode scanner flags healthy containers as UNHEALTHY due to system resource limits
Previously, background container scanners in Ozone incorrectly marked containers as UNHEALTHY when they encountered system resource issues, such as file-descriptor exhaustion resulting in Too many open files errors, instead of actual data corruption. This behavior led to false positives because healthy containers were flagged due to temporary system limits.

This issue is now resolved, and the logic is updated to catch these resource-related exceptions, such as FileNotFoundException or FileSystemException, ensuring that containers remain in their current state if the scanner cannot perform its check due to system constraints.

Apache Issue: HDDS-15150
CDPD-102273: Unbounded GRPC server connections lead to DataNode file handler exhaustion
Previously, no limit existed on the number of network connections created by the DataNode GRPC server. This led to too many open files errors, because TCP connections consumed the majority of available file handlers, potentially causing DataNode instability.

This issue is now resolved. The GrpcConnectionLimitFilter connection limit filter (which defaults to a maximum of 5,000 connections) and an OS backlog now manage burst connections, ensuring that sufficient file handlers remain available for files and databases.

Apache Issue: HDDS-15149
CDPD-105094: Erasure-coded writes fail with ArrayIndexOutOfBoundsException due to race condition
Previously, a race condition in the CoderUtil shared cache caused erasure-coded (EC) writes to fail with an ArrayIndexOutOfBoundsException error when concurrent operations requested zero-filled buffers of different sizes.

This issue is now resolved. The shared empty chunk cache is updated to handle concurrent access safely.

Apache Issue: HADOOP-19900
CDPD-101970: Ozone client hangs during writeStateMachineData failures
Previously, when a failure occurred during the writeStateMachineData operation on an Ozone DataNode, such as a disk space issue or other exceptions, the client hung for an extended period while the leader node repeatedly retried the operation until the pipeline closed. This resulted in significant delays before the operation was finally marked as a failure.

This issue is now resolved by optimizing failure handling to reduce the client waiting time to a few seconds in most failure scenarios.

Apache Issue: HDDS-15122
CDPD-103013: Delay in safemode exit on follower causing intermittent behavior leading to leader switch test failure
Previously, Ozone Storage Container Manager (SCM) follower nodes could remain stuck in safe mode after a restart. This occurred because the ContainerSafeModeRule class refreshed its container list and potentially added newly created containers that DataNodes (DNs) had already registered. Consequently, the system could not meet the 99% container reporting threshold, delaying the safe mode exit until a new Ratis transaction was applied, causing intermittent failures in leader switch tests and impacting small clusters.

This issue is now resolved by ensuring the safe mode rule maintains the container list initialized at startup and only removes deleted containers, correctly updating the threshold and cutoff values.

Apache Issue: HDDS-15238
CDPD-105034: Missing scanner instance owner check in HBase Thrift and REST services
Previously, HBase Thrift and REST did not verify scanner ownership on fetch and close operations in the scan workflow (open, fetch, close). An authenticated user could read rows from scanners opened by other users or close scanners belonging to other users. For more information, see CVE-2026-49326.

This issue is now fixed. The services verify that the effective user matches the scanner owner before allowing fetch or close operations.

Apache Jira: HBASE-30183

CDPD-106891: Avoid using reflection to initialize ServerCacheFactory in Apache Phoenix
Previously, the Phoenix query server could instantiate a ServerCacheFactory implementation using reflection from a class name supplied in the server-caching RPC request. This allowed a caller to load disallowed classes on the RegionServer.

This issue is now fixed. The server creates only the supported factory types (HashCacheFactory and IndexMetaDataCacheFactory) and rejects other class names.

Apache Jira: PHOENIX-7933

CDPD-106986: Node.js security vulnerabilities in Streams Messaging Manager UI on RHEL 8

Previously, the Node.js version bundled with Streams Messaging Manager UI on RHEL 8 contained security vulnerabilities, specifically CVE-2025-23083, CVE-2025-23084, and CVE-2025-23085.

This issue is now resolved. Streams Messaging Manager UI on RHEL 8 now includes an updated Node.js version that remediates these CVEs.

CDPD-80868: Basic search API returns wrong response for attribute-based search with long values
Previously, basic search using attribute filters, such as qualifiedName, returned no results when the filter values exceeded the default maximum token length of 255 characters configured in Solr. This issue affected the startsWith, endsWith, and contains operators.
This issue is now resolved. For string filters longer than the configured token length, the search now falls back to evaluate the criterion on JanusGraph instead of Solr.

Apache Jira: ATLAS-5032

CDPD-69317: Export/Import API: Import fails when a tag attribute type changes between exports
Previously, Replication Manager export or import failed when a classification (tag) attribute type changed between exports. For example, after a tag was deleted and recreated on the source with the same name but a different attribute type (such as string to double), incremental import failed with the INVALID_IMPORT_ATTRIBUTE_TYPE_CHANGED error or a JanusGraph schema error because the target retained the old typedef and property key types.
This issue is now fixed. The import process now treats a changed attribute type as a typedef update and recreates the JanusGraph property key so that classification values persist with the new data type.

Apache Jira: ATLAS-5310

CDPD-101416: Schema tab does not display column records on the entity detail page
Previously, on the entity detail page, the Schema tab did not list schema rows (for example, columns). The table was displayed empty even when records were expected to be present.
This issue is now fixed. The Schema tab lists schema rows with correct columns and pagination, consistent with relationship search and referred entities.

Apache Jira: ATLAS-5269

CDPD-103032: Business metadata with an array attribute type is slow to delete
Previously, when you deleted a business metadata typedef that included an attribute with the array data type, Apache Atlas processing was slow because it scanned Solr for entity references even though array attribute values are not Solr-indexed.

This issue is now resolved. Deletion now uses graph-based existence checks, which significantly reduces deletion time on large catalogs. For business metadata typedef deletes through the REST API, you can use the optional force query parameter on DELETE /api/atlas/v2/types/typedefs and DELETE /api/atlas/v2/types/typedef/name/{typeName} REST API endpoints. Setting force=true applies only to business metadata typedefs and skips reference validation. Other typedef categories retain the previous validation behavior.

Apache Jira: ATLAS-4988

CDPD-103757: Ranger does not validate the doAsUser parameter for impersonation permission
Previously, the RangerJwtAuthHandler class accepted the doAsUser value directly from the incoming request and used it to establish the authenticated user identity without performing any authorization check. This behavior could allow unauthorized impersonation.

This issue is now fixed. Ranger validates that the requesting user has impersonation permission before honoring the doAsUser parameter.

CDPD-100919: Ranger KMS does not check for master key existence on GCP before attempting to create one
Previously, when Ranger KMS was configured with a GCP HSM backend, restarting the Ranger KMS service after revoking the cloudkms.cryptoKeys.create permission would fail with a PERMISSION_DENIED error. This occurred because Ranger KMS attempted to create the master key on every service restart, regardless of whether the key already existed in GCP. Since the create permission is typically revoked after the initial one-time master key creation, subsequent restarts would fail.

This issue is now fixed. Ranger KMS correctly checks for the existence of the master key in GCP before attempting to create a new one, allowing service restarts to succeed even after the create permission is revoked.

KT-7622: Security vulnerabilities in Key Trustee Server embedded PostgreSQL

Previously, the embedded PostgreSQL database bundled with Key Trustee Server contained security vulnerabilities, including CVE-2025-12817, CVE-2025-12818, CVE-2026-2003, CVE-2026-2004, CVE-2026-2005, and CVE-2026-2006. These vulnerabilities included risks such as heap buffer overflows in pgcrypto, missing input validation in intarray, and multibyte character validation flaws that could lead to arbitrary code execution.

This issue is now resolved. Key Trustee Server now bundles PostgreSQL version 14.21, which remediates these security vulnerabilities.

COMPX-26377 : Issue in YARN Queue Manager UI during cluster restart
Previously, YARN Queue Manager UI stopped responding when loading the screen after a cluster restart.

This issue is now resolved and the interface now loads correctly on the first request without requiring a page refresh.

CDPD-102464: ZooKeeper configuration properties logged without redaction
Previously, ZooKeeper configuration properties were recorded in plain text, potentially exposing sensitive data within the logs.

This issue is now resolved by implementing a log redaction method to ensure that all sensitive values are appropriately masked during the logging process.

CDPD-102463: Reverse DNS lookup in TLS client and server
Previously, ZooKeeper performed reverse DNS lookups during TLS handshakes in both the client and server, which introduced a high security vulnerability.
This issue is now resolved. Reverse DNS lookup is now disabled in the ZooKeeper TLS client and server.
Apache Jira - ZOOKEEPER-4986
CDPD-109656: Sqoop data imports do not support read-only databases
Previously, Sqoop failed during data import operations on read-only databases (such as Netezza) because it attempted to run an unsupported commit() call, triggering exceptions and task failures. This issue is now fixed.
CDPD-104785: Arbitrary file read on Hue hosts through the localfile importer feature
Previously, a local file inclusion vulnerability existed on the Hue host type through the localfile importer feature. This occurred because two indexer API endpoints accepted unvalidated paths, allowing unauthorized file reads through the path parameter. This issue is now fixed.
CDPD-68779: Error while browsing S3 buckets or ADLS containers from the left-assist panel
Previously, when attempting to browse S3 buckets or ADLS containers from the left-assist panel in Hue without the required permissions, the system displayed the Failed to retrieve buckets:1:0: syntax generic error message.

This issue is now fixed. When RAZ is enabled, the left-assist panel now opens the home directory if the required permissions are available. In non-RAZ environments, the left-assist panel opens the path configured in the Hue configuration file.

CDPD-90751: Reverse tabnabbing security vulnerability in external links
Previously, external links using the target="_blank" link attribute without the rel="noopener" or rel="noreferrer" property values created a reverse tabnabbing security vulnerability. This issue is now fixed.
Query failure with combined IN or EXISTS subqueries and correlated conditions
Previously, a query failed during compilation with an IndexOutOfBoundsException error when its WHERE clause combined an IN or EXISTS subquery with a correlated condition.
This issue is now resolved, and these queries run successfully.

Apache Jira: HIVE-29688

Common Vulnerabilities and Exposures (CVE) that is fixed in this CHF: