Known Issues in Apache Knox

Known issues and technical limitations for Apache Knox are addressed in Cloudera Runtime 7.3.3.

Known Issues identified in Cloudera Runtime 7.3.3

CDPD-126571: JDBCTokenStateService fails to initialize on PostgreSQL when multiple Knox instances start at the same time
7.3.3
In a Knox high availability deployment that uses PostgreSQL, JDBCTokenStateService can fail to initialize when multiple Knox Gateway instances start at the same time against an empty database. Startup fails with an error similar to the following:
PSQLException: ERROR: duplicate key value violates unique constraint "pg_type_typname_nsp_index"
Detail: Key (typname, typnamespace)=(knox_tokens, 2200) already exists.
Restart the Knox service on the host where startup failed. The KNOX_TOKENS and KNOX_TOKEN_METADATA tables already exist, so the service starts normally. To avoid the issue on the first startup, start Knox on one host at a time, or create the KNOX_TOKENS and KNOX_TOKEN_METADATA tables before you start Knox.
Apache JIRA: KNOX-3536

Known Issues identified before Cloudera Runtime 7.3.3

Known issues identified before Cloudera Runtime 7.3.3 include only unresolved issues from previous releases that continue to affect the Cloudera Runtime 7.3.3 base release.

CDPD-104007: Tokens on Knox Token Management page might intermittently fail to display
7.3.3, 7.3.2.100 CHF 1 and higher
Tokens created under the Knox Token Management page might intermittently fail to display. The issue occurs randomly upon refreshing the page, where the underlying API call to retrieve user tokens is not triggered, causing the page to appear empty.
Refresh the page again until the tokens are displayed.
OPSAPS-76713: Knox crashes under high concurrent load when using PAM authentication on RHEL 9 or Rocky Linux 9
7.3.3, 7.3.2 and its SPs and CHFs
On RHEL 9 or Rocky Linux 9 versions 9.1 through 9.7, a concurrency bug in libeconf 0.4.1 causes libpam to crash under high concurrent load. This affects Knox when the cdp-proxy-api topology uses PAM authentication. When Knox handles multiple concurrent requests, libpam might crash, causing Knox service to crash and requests to the Knox cdp-proxy-api topology to fail.
Use AD/LDAP authentication instead of PAM authentication for the cdp-proxy-api topology. In Cloudera Manager, go to the Knox service, select Configuration, and configure the gateway_api_authentication_provider parameter to use AD/LDAP authentication. For configuration instructions, see Configure Apache Knox authentication for AD/LDAP.
CDPD-76294: Knox service can not be started in a large size Private Cloud Base Cloudera Manager cluster
7.3.3, 7.3.2 and its SPs and CHFs, 7.3.1.100 through 7.3.1.700
For a large size Private Cloud Base Cloudera Manager cluster installed with Cloudera Runtime 7.3.1.0, you might face the problem that Knox can not be started with the following error message:
Wait Until Knox Gateway Can Serve Requests failed on Knox Gateway
Increase the Knox configuration parameter Knox Gateway Initial/Max Heapsize from 1 GiB to 2 GiB or 4 GiB, depending on the cluster size. Then save changes and run Restart Stale Services. After these steps, the Knox service can be started.
CDPD-71751: Creation of alias from the Cloudera Manager UI fails on FIPS
7.3.3, 7.3.2 and its SPs and CHFs, 7.3.1 and its SPs and CHFs, 7.1.9 SP1 and its CHFs
Users attempting to create aliases through the Cloudera Manager UI face issues in FIPS.
The alias(es) can be created using the Knox CLI:
  1. ssh to Knox host.
  2. Export these directories: export KNOX_GATEWAY_DATA_DIR="/var/lib/knox/gateway/data"; export KNOX_GATEWAY_CONF_DIR="/var/lib/knox/gateway/conf"
  3. Set the FIPS-specific options for the Knox CLI:
    export KNOX_CLI_MEM_OPTS="--add-exports=java.base/sun.security.provider=bctls --add-exports=java.base/sun.security.provider=com.safelogic.cryptocomply.fips.core --add-modules=com.safelogic.cryptocomply.fips.core --add-modules=bctls --module-path=<BCTLS_JARS_DIR> -Dcom.safelogic.cryptocomply.fips.approved_only=true"
    <BCTLS_JARS_DIR> is the directory containing the SafeLogic bctls and fips core jar files.
  4. Run the following command to create the alias: /opt/cloudera/parcels/CDH/lib/knox/bin/knoxcli.sh create-alias <ALIAS_NAME> <ALIAS_VALUE>
  5. Verify the addition using /opt/cloudera/parcels/CDH/lib/knox/bin/knoxcli.sh list-alias.

For HA deployments, users must do it on every Knox host (whereas the Save Alias command applies the change to all hosts automatically).

CDPD-71305: Concurrent impala shell connection failure
7.3.3, 7.3.2 and its SPs and CHFs, 7.3.1 and its SPs and CHFs, 7.1.9 SP1 and its CHFs
If a user makes a concurrent impala-shell connection through Knox, then the connection fails.
Use only one Knox role.
CDPD-3125: Logging out of Atlas does not manage the external authentication
7.3.3, 7.3.2 and its SPs and CHFs, 7.3.1 and its SPs and CHFs, 7.2.18, 7.1.9
At this time, Atlas does not communicate a log-out event with the external authentication management, Apache Knox. When you log out of Atlas, you can still open the instance of Atlas from the same web browser without re-authentication.
To prevent additional access to Atlas, close all browser windows and exit the browser.
CDPD-28431: Intermittent errors could be potentially encountered when Impala UI is accessed from multiple Knox nodes
7.3.3, 7.3.2 and its SPs and CHFs, 7.3.1 and its SPs and CHFs, 7.1.9
You must use a single Knox node to access Impala UI.
CDPD-22785: Improvements and issues need to be addressed in convert-topology knox cli command
7.3.3, 7.3.2 and its SPs and CHFs, 7.3.1 and its SPs and CHFs, 7.1.9
None.
Knox issue with JDK version
7.3.3, 7.3.2 and its SPs and CHFs, 7.3.1 and its SPs and CHFs, 7.1.9
jdk-1.8.0_391 is not supported.
Cloudera recommends using Cloudera supported JDKs.
CDPD-84236: Token generated by one Knox host fails with Unknown token error on another Knox host in Data Engineering High Availability clusters
7.3.3, 7.3.2 and its SPs and CHFs, 7.3.1.400 through 7.3.1.700
In Data Engineering High Availability clusters, a token generated by one Knox host may fail with an Unknown token error when accessed through another Knox host. This issue occurs due to a race condition in the PostgreSQL database, which prevents one of the Knox instances from properly initializing its configured token state service.
Restart Knox on all hosts.