4. Known Issues

Ambari 1.7.0 has the following known issues, scheduled for resolution in a future release. Please work around the following issues with Ambari 1.7.0:

 

Table 3.3. Ambari 1.7.0 Known Issues

Issue Tracking #

Title

Problem Description

Solution

BUG-33118

YARN App Timeline Server does not start after upgrade to Ambari 1.7.0

After upgrading to Ambari 1.7.0, the YARN App Timeline Server (ATS) does not start when using HDP 2.1.2 (or earlier) Stack. THe ATS server command was changed in HDP 2.1.3 or later and therefore, if you are using HDP Stack 2.1.2 (or earlier), ATS will fail to start with the following error:

Fail: Execution of 'ls /var/run/hadoop-yarn/yarn/yarn-yarn-timelineserver.pid` >dev/null
                           2>&1 && ps `cat /var/run/hadoop-yarn/yarn/yarn-yarn-timelineserver.pid` >/dev/null
                           2>&1' returned 1.

Please contact Hortonworks Support and reference BUG-33118.

BUG-29221

Tez io.sort should not be set above 2GB.

If Tez tez.runtime.io.sort.mb is set above 2GB, Tez jobs will fail.

Adjust the Tez configuration property tez.runtime.io.sort.mb to be 2GB. Ambari will warn that this value is below the recommended value. Proceed anyway and save the value.

BUG-28651

Ambari user+group sync fails with Active Directory

When configuring Ambari for Active Directory authentication, user and group import fails. This is due to an incorrect mapping of uid to the username attribute in the Ambari import/sync process. An error message similar to the following is printed to ambari-server.log: 17:20:38,375 WARN [pool-1-thread-2] AmbariLdapDataPopulator:560 - Ignoring LDAP user cn=testUser,ou=Ambari,dc=SME, dc=support,dc=com as it doesn't have required attributes uid and sAMAccountName

Please contact Hortonworks Support and reference BUG-28651 to obtain a patch.

BUG-28245

Slider View: Uncaught error during recreate STORM app with the same name.

When you delete, then recreate a Storm application using Slider view, browser returns uncaught error 405, which results in the app not showing in the Slider View.

Refresh the browser and the app will display.

BUG-28235

Unable to install Falcon Service with HDP 2.1 stack.

Unable to install Falcon with HDP 2.1 stack due to required properties. When using the HDP 2.1 Stack and attempting to install Falcon (either during initial cluster Install or via Add Service), Falcon prompts for two required properties. You are unable to proceed without entering values for the properties.

Enter the following values for the properties to proceed with the Falcon install: *.falcon.graph.blueprints.graph= com.thinkaurelius.titan.core.TitanFactory *.falcon.graph.storage.backend= berkeleyje

BUG-28234

Hive filesystem cache must be disabled to prevent memory leak

Hive Server 2 out of memory due to FileSystem.CACHE leak, when operating in un-secure mode.

Disable file system caches by setting the following parameters to true in the hive configuration (hive-site.xml): fs.hdfs.impl.disable.cache, default false. fs.file.impl.disable.cache, default false.

BUG-28203

Stop Services fails while running Enable Kerberos wizard.

When running the Enable Kerberos wizard, the Stop Service step fails due to a RegionServer failing to stop. This is due to attempting to stop RegionServer after the NameNode is already stopped.

Perform a hard kill on the RegionServer and delete the RegionServer PID file. Restart the Enable Kerberos Wizard and the Stop Services step should succeed.

BUG-28200

WebHCat webhcat-site templeton.hive.properties is set incorrectly.

After installing Ambari 1.7.0, in the WebHCat webhcat-site configuration, the templeton.hive.properties property is set with spaces and the incorrect hive.metastore.uris value (the port is incorrect). hive.metastore.local=false, hive.metastore.uris=thrift://hdp-22.harry.inc:9983, hive.metastore.sasl.enabled=false, hive.metastore.execute.setugi=true

Remove the spaces and set the hive.metastore.uris property correctly. Be sure that value matches the hive.metastore.uris property URL.

BUG-28199

Run ambari-server using non-root causes issues with view instances.

If you have setup Ambari Server to run as a sudoer user (and not as “root”), when attempting to access a view instance you receive “Velocity could not be initialized!” error.

The views framework processes instances using Velocity templates and when the Ambari Server is running as non-root, the velocity log is not writable by default. You need to modify the permissions and ownership for /velocity.log so Ambari Server can access the file.

BUG-28151

NameNode is not started error when running NameNode HA Wizard.

When running the NameNode HA wizard, you receive a “NameNode is not started. Please quit wizard and start NameNode first” error message.

Close the NameNode HA wizard, restart the NameNode and restart the wizard.

BUG-27806

Enabling Yarn RM HA through Ambari should enable work preserving RM restart

Work Preserving Restart preserves in-flight job task work across restarts of the ResourceManager or NodeManager. Work Preserving Restart is not enabled by default in HDP installation when Resource Manager HA is enabled.

To configure Work Preserving Restart, configure the following in yarn-site.xml after enabling ResourceManager HA: yarn.nodemanager.recovery. enabled = true yarn.resource manager.recovery.enabled = true yarn.resourcemanager.work- preserving-recovery.enabled = true Configure a local directory for the NodeManager to store recovery state in: yarn.nodemanager.recovery.dir = /var/log/hadoop-yarn/nodema nager/recovery-state

BUG-27492

Add Service create multiple Service Configuration Versions without changes.

After adding a service using the Add Service wizard, the HDFS Service includes multiple Service Configuration Versions without a note or changes.

These configuration versions are benign.

BUG-27434

HDP 1.3 Stack “Scheduling Info” quick link is not available.

When using HDP 1.3 Stack and configuring Kerberos, the “Scheduling Info” quick link under the MapReduce service is unavailable.

The link is incorrect.

BUG-27419

HDP 1.3 Stack has redundant 'DECOMMISSION' option for RegionServer component.

When using HDP 1.3 Stack, the RegionServer component Actions menu includes two Decommission options.

The options are redundant.

BUG-27380

Host page size does not preserve after logout.

On the Hosts page, changing the page size is not preserved between logins and reverts to 10.

On each new login, you will have to reset the page size to the desired value.

BUG-27212

Installing a cluster (via Install Wizard or Blueprints) with Hive and an existing MySQL database existing on the same host as the Hive Metastore server can cause installation conflicts.

When using an existing MySQL database for Hive Metastore, if you have the existing MySQL database on the same server as the Hive Metastore server, during install, Ambari might fail to install Hive due to conflicts with mysql.

Do not install the Hive Metastore component on the same host with the existing MySQL database.

BUG-27129

Blueprint is installing MySQL even when using "Existing MySQL" Hive option.

Unable to install single-node cluster with existing MySQL database via Blueprint.

Cluster install will fail on a single-node cluster when using an existing MySQL database on that same host for both Ambari and Hive Metastore. The existing MySQL database should not be located on the single-node cluster host, or Ambari should use the default embedded PostgreSQL database.

BUG-26650

Problems with Oozie and MapReduce checks after upgrade to Ambari 1.7.0.

After upgrading from Ambari 1.6.0 to 1.7.0 with HDP 2.1 and Kerberos, MapReduce checks fail, which causes other checks (like Oozie) to also fail.

If you are running the following configuration, you must modify the YARN property to confirm ATS is not enabled. Start environment is Ambari 1.6.0 + HDP 2.1 Stack + Kerberos enabled. Before upgrade to Ambari 1.7.0, set the following property: 1. Browse to Services > YARN > Configs > yarn-site 2. Set yarn.timeline-service.enabled = false 3. Save the configuration. Restart YARN, Oozie, and MapReduce.

BUG-25056

HDFS metrics do not displaying after adding Ganglia.

If you add Ganglia post-cluster install (via Add Service), metrics do not show up in the other Services and no restart indicators are shown.

Restart the other Stack services for the Ganglia metrics to start displaying.

BUG-24902

Setting cluster name to more than 100 chars hangs ambari-server.

In the Administration interface, if you rename your cluster to something > 100 chars, the Ambari Server hangs with 500 error.

Restart the Ambari Server to clear the error.

BUG-18369

ATS start may fail due to ClassNotFoundException

The Application Timeline Server (ATS) component of YARN configuration property changed with the HDP 2.1.3 release. If you install a patch release of HDP with Ambari 1.7.0, you must change a YARN configuration property: Running Ambari 1.7.0 and HDP 2.2 without changing this configuration may cause ATS start to fail with the following error message: Fail: Execution of 'ls /var/run/hadoop-yarn/yarn/yarn-yarn-historyserver.pid>/dev/null 2>&1 && ps 'cat /var/run/hadoop-yarn/yarn/yarn-yarn-historyserver.pid'>/dev/null 2>&1' returned 1.

Browse to Ambari Web > Services > YARN Configs and make the following modification: If you install Ambari 1.7.0 and use HDP 2.1.2 or 2.1.1: yarn.timeline-service.store- class=org.apache.hadoop.yarn. server.applicationhistoryservice. timeline.LeveldbTimelineStore If you install and use HDP 2.1.3 and higher, or HDP 2.2: yarn.timeline-service.store- class=org.apache.hadoop.yarn. server.timeline.LeveldbTimeline Store