Known issues and technical limitations for HBase are addressed in Cloudera Runtime 7.3.3.
Known issues identified in Cloudera Runtime 7.3.3
- OPSAPS-81682: HBase clients cannot connect when RPC mTLS client
authentication is required
- 7.3.3
- When AutoTLS is enabled and the HBase RPC mTLS client auth
mode property is set to Required (NEED), client TLS
keystores are not generated. As a result, HBase clients, such as HBase shell, cannot connect
to HBase Master, and commands fail with the Received fatal alert:
bad_certificate error.
- Apply either of the following configuration changes:
- Change the HBase RPC mTLS client auth mode
(hbase.server.netty.tls.client.auth.mode) property from
Required (NEED) to Disabled
(NONE) or Optional (WANT).
- Enable the Master / Region Server TLS Plaintext Enabled
(hbase.server.netty.tls.supportplaintext) property.
Known issues identified before Cloudera Runtime 7.3.3
- CDPD-109095: Client-side TLS hostname validation issue in Apache
Thrift (libthrift)
- 7.3.2.20000 SP2
- A TLS hostname validation issue exists in libthrift versions
lower than 0.23.0. The issue does not allow remote compromise of your CDP HBase server,
but external client applications that connect to the HBase Thrift gateway over TLS might
be vulnerable to Man-in-the-Middle (MitM) traffic interception if an attacker redirects
network traffic.
- Update libthrift to 0.23.0 or higher versions in your
client applications. Version 0.23.0 is fully wire-compatible with the CDP HBase Thrift
server. If an immediate client library update is not possible, perform one of the
following actions:
- Enforce Kerberos (SASL) authentication on the HBase Thrift gateway by using
Cloudera Manager.
- Configure client socket settings to explicitly perform TLS hostname
verification.
For more information, see CVE-2026-43869.
- CDPD-119707: HBase replication between TLS-enabled Cloudera Operational Database clusters fails due to hostname verification
- 7.3.2.10000 SP1
- When you set up reverse HBase replication between two
TLS-enabled Cloudera Operational Database clusters, replication fails if reverse
DNS resolution is not configured between the clusters. During mutual TLS (mTLS)
authentication, the source cluster sends its certificate, which contains the hostname.
The target cluster validates the certificate against its truststore and then verifies
that the client connection IP address matches the hostname in the certificate. Because
reverse DNS resolution does not work between separate Cloudera Operational Database clusters by default, this verification check fails and the connection is
rejected.
This issue occurs when all of the following conditions are true:
- Both clusters have TLS enabled.
- Hostname verification is enabled
(
hbase.client.netty.tls.verify.server.hostname and
hbase.server.netty.tls.verify.client.hostname are set to
true).
hbase.server.netty.tls.client.auth.mode is set to
WANT or NEED.
- TLS certificates are issued without an IP Subject Alternative Name (SAN).
- Reverse DNS resolution is not configured between the clusters.
- Mutual TLS trust is established between the clusters.
These are the default settings for a fresh installation of Cloudera Runtime 7.3.2.10000.
- To bypass the client hostname verification check, add the
hbase.server.netty.tls.verify.client.hostname = false property to the
HBase Service Advanced Configuration Snippet (Safety Valve) for
hbase-site.xml using Cloudera Manager.