Cloudera Docs

Step 9: Complete Post-Upgrade steps for upgrades to CDP Private Cloud Base

Steps to perform after upgrading a cluster.

Loading Filters ... 7.0.3 7.1.1 7.1.2 7.1.3 7.1.4 7.2.4 7.3.1 7.4.4 7.5.1 7.6.1 7.6.7 7.7.1 7.7.3 7.11.3 7.1.7.2000 7.1.8 7.1.7.1000 7.1.7 7.1.6 7.1.5 7.1.4 7.1.3 7.1.2 7.1.1 7.1.9

Some components might require additional steps after you complete the upgrade to CDP Private Cloud Base:As a post-upgrade step, you must check that location in HDFS to ensure that it is the latest version, and run the following if it does not:
hdfs dfs -ls /user/yarn/mapreduce/mr-framework/
If only the old version's tar file exists there, such as:
/user/yarn/mapreduce/mr-framework/3.1.1.7.1.7.78-12-mr-framework.tar.gz
Then perform the following: Cloudera Manager > Yarn > Actions > Install YARN MapReduce Framework JARs.You can clear out the old tar file after this is done for keeping it clean.
Tez MR Framework JARs and Oozie Sharelib entities must be examined in the similar manner. Check that location in HDFS to ensure that it is the latest version, and run the following if it does not:
  1. Cloudera Manager > Yarn > Actions > Install YARN MapReduce Framework JARs
  2. Cloudera Manager > Tez > Actions > Upload Tez tar file to HDFS
  3. Cloudera Manager > Oozie > Actions > Install Oozie ShareLib
Clear any old files out of these directories: 
  1. hdfs dfs -ls -R /user/yarn/mapreduce/mr-framework/
  2. hdfs dfs -ls -R /user/tez/
  3. hdfs dfs -ls /user/oozie/share/lib/

Kafka – clear protocol and log format versions

  1. Remove the following properties from the Kafka Broker Advanced Configuration Snippet (Safety Valve) for kafka.properties configuration property.
    • inter.broker.protocol.version
    • log.message.format.version
  2. Save your changes.
  3. Perform a rolling restart:
    1. Select the Kafka service.
    2. Click Actions > Rolling Restart.
    3. In the pop-up dialog box, select the options you want and click Rolling Restart.
    4. Click Close once the command has finished.

Kafka – run the Create Kafka Log Directory Free Capacity Check command

In Cloudera Runtime 7.1.9, a new Cloudera Manager trigger called Broker Log Directory Free Capacity Check is introduced. This trigger fires if the capacity of any log directory falls below 10%. The trigger is automatically created on new clusters, but must be created on upgraded clusters by running the Create Kafka Log Directory Free Capacity Check command. Complete the following to create the trigger:
  1. In Cloudera Manager, select the Kafka service.
  2. Click Actions > > Create Kafka Log Directory Free Capacity Check.

Once created, the trigger is available on the Kafka service > Status page. Note that the Status page also contains a new chart which is introduced in 7.1.9. The chart is called Log Directory Free Capacity. It shows the capacity of each Kafka Broker log directory.

Kafka producer – disable idempotence or update Ranger policies

Kafka producer applications upgraded to Kafka 3.0.1, 3.1.1, or all versions after 3.1.1 have idempotence enabled by default. Idempotent producers must have Idempotent Write permission set on the cluster resource in Ranger. As a result, If you upgraded your producers and use Ranger to provide authorization for the Kafka service, you must do either of the following after upgrading your producers. If configuration is not done, the producers fail to initialize due to an authorization failure.
  • Explicitly disable idempotence for your producers. This can be done by setting enable.idempotence to false for your producer applications. For more information, see Producer Configs in the Apache Kafka documentation.
  • Update your policies in Ranger and ensure that producers have Idempotent Write permission on the cluster resource. For more information, see Configuring resource-based policies.

Kafka Connect – resolve ClassNotFoundException

Upgrading a Kafka Connect cluster can fail with the following ClassNotFoundException: 
Error: ClassNotFoundException exception occurred: com.cloudera.dim.kafka.config.provider.CmAgentConfigProvider
exception occurred: com.cloudera.dim.kafka.config.provider.CmAgentConfigProvider
This is due to the Cloudera Manager agent not being able to properly create the symlinks and alternatives when performing an upgrade. The issue also occurs when a custom path is set for plugin.path in Cloudera Manager, which does not contain the expected plugin JARs. Complete the following steps to resolve the issue:
  1. Manually install missing symlinks or alternatives relevant to kafka-connect by checking cloudera-scm-agent.log.
  2. Ensure that the kafka-connect libraries are present in /var/lib/kafka or change the plugin.path to a value where these libraries are present.

    If required, manually copy the JARs to the location configured in plugin.path.

  3. If the issue persists, set the plugin.path to:
    /opt/cloudera/parcels/CDH-[***VERSION***]/lib/kafka_connect_ext/libs
    This is the location where kafka-connect libraries are set when installing parcels.

Kafka Connect – configure Kerberos ATL rules

Kafka Connect now uses the cluster-wide Kerberos auth-to-local (ATL) rules by default. A new configuration property called Kafka Connect SPNEGO Auth To Local Rules is introduced. This property is used to manually specify the ATL rules. During an upgrade, the property is set to DEFAULT to ensure backward compatibility. If you want to use the cluster-wide Kerberos ATL rules for custom Kerberos principals, you need to clear the existing value from the Kafka Connect SPNEGO Auth To Local Rules property in Cloudera Manager > Kafka > Configuration.

Schema Registry - create the Ranger Schema Registry Plugin Audit Directory

The Ranger Schema Registry Plugin Audit Directory is not created automatically when upgrading a cluster. This causes the Schema Registry service to encounter an issue. As a result, following an upgrade, you must manually initiate the command in Cloudera Manager that creates the audit directory.
  1. In Cloudera Manager, select the Schema Registry service.
  2. Click Actions > Create Ranger Schema Registry Plugin Audit Directory.

Schema Registry – configure Kerberos Name Rules

The default value for Schema Registry Kerberos Name Rules is changed. Schema Registry now automatically applies the cluster-wide auth-to-local (ATL) rules by default. However, the previous value of the Schema Registry Kerberos Name Rules property is preserved during an upgrade. You must manually clear the property following an upgrade to transition to the new default value in Cloudera Manager > Schema Registry > Configuration when using Custom Kerberos Principal.

Streams Messaging Manager – update the replication factor of internal topics

In Cloudera Runtime 7.1.9, the default replication factor of the __smm* internal SMM topics is increased to 3. The replication factor of these topics is not updated when you upgrade a cluster. As a result, Cloudera recommends that you increase the replication factor of these topics to 3 with the kafka-reassign-partitions tool following the upgrade. For more information, see kafka-reassign-partitions.

Streams Messaging Manager – reconnect SMM with SRM

Following a successful upgrade, if you want to use SMM to monitor SRM replications, you must reconnect the two services. This is done by enabling the STREAMS_REPLICATION_MANAGER Service SMM property which is disabled by default.
  1. In Cloudera Manager, select the SMM service.
  2. Go to Configuration.
  3. Find and enable the STREAMS_REPLICATION_MANAGER Service property.
  4. Click Save Changes.
  5. Restart the service.

Streams Messaging Manager – configure Kerberos Name Rules

The default value for Kerberos Name Rules is changed. SMM now automatically applies the cluster-wide auth-to-local (ATL) rules by default. However, the previous value of the Kerberos Name Rules property is preserved during an upgrade. You must manually clear the property following an upgrade to transition to the new default value in Cloudera Manager > SMM > Configuration when using Custom Kerberos Principal.

Streams Messaging Manager – update Prometheus configuration

The default port that the Kafka Connect metrics reporter uses as well as the property that configures the port are changed. If you are using Prometheus as the metrics store for SMM, you must update your Prometheus configuration to include the new port.

Complete the following:
  1. In Cloudera Manager, select the Kafka service.
  2. Go to Configuration.
  3. Take note of the values of Secure Jetty Metrics Port and Jetty Metrics Port properties.

    Which property value you take note of depends on whether Kafka Connect is secured using TLS/SSL. Kafka Connect is secured with TLS/SSL if Enable TLS/SSL for Kafka Connect is selected. Take note of Jetty Metrics Port if Kafka Connect is not secured. Take note of Secure Jetty Metrics Port if Kafka Connect is secured.

  4. If Enable Basic Authentication for Metrics Reporter is selected take note of the following properties:
    • Jetty Metrics User Name
    • Jetty Metrics Password
  5. Update the kafka_connect job section in your prometheus.yml configuration file with the new port number. If Basic Authentication is enabled, add the username and password as well. For example:
    - job_name: 'kafka_connect'

    metrics_path: '/api/prometheus-metrics'

    basic_auth:
      username: 'email@username.me'
      password: 'password'

    static_configs:
    - targets: ['my-cluster-1.com:28087','my-cluster-2.com:28087','my-cluster-3.com:28087']
  6. Reload your Prometheus configuration.

    For more information, see Configuration in the Prometheus documentation.

Streams Replication Manager – disable legacy protocol and data format

During the upgrade SRM uses the legacy versions of its intra cluster protocol and internal changelog data format to ensure backward compatibility. After a successful upgrade, you must configure SRM to use the new protocol and data format. Otherwise, some features will not function.
  1. In Cloudera Manager select the Streams Replication Manager service.
  2. Go to configuration and clear the following properties:
    • Use Legacy Intra Cluster Host Name Format
    • Use Legacy Internal Changelog Data Format
  3. Restart the SRM service.

Streams Replication Manager – reconfigure the Log Format property

The value set in the Log Format property is cleared during the upgrade. If you customized the log format using this property, its value must be manually reconfigured following the upgrade.
  1. In Cloudera Manager select the SRM service.
  2. Go to Configuration and configure the Log Format property.
  3. Restart the SRM service.

Streams Replication Manager – export consumer group offsets

In Cloudera Runtime 7.1.8 and higher, the separator configured with replication.policy.separator in Streams Replication Manager applies to the names of the offsets-syncs and checkpoints internal topics if the default replication (DefaultReplicationPolicy) policy is in use.

If you have been using a custom separator, when SRM is restarted after the upgrade, a new set of internal topics are automatically created using the separator that is configured. These topics, however, will be empty and are only be repopulated with data after replication restarts. If you want to export translated consumer group offsets after the upgrade but before replication starts, you must export the translated consumer group offsets from the old topics as the data is not yet available in the new topics. This is done by specifically instructing the srm-control tool to use the default separator, which is the dot (.). For example:
srm-control offsets --config [***CONFIG FILE***] --source [SOURCE_CLUSTER] --target [TARGET_CLUSTER] --group [GROUP1] --export > out.csv
Where the [***CONFIG FILE***] specified in --config contains the following configuration entry
replication.policy.separator=.

Streams Replication Manager – delete unused internal topics

In Cloudera Runtime 7.1.9 and higher, SRM uses a new version of its internal Kafka Streams application. The topics and data of the old application (the one used before the upgrade) are not deleted automatically during the upgrade. Run the following commands to reset the state of the now unused Kafka Streams applications and to delete unused internal topics. Run the following commands on the hosts of the SRM Service role.
kafka-streams-application-reset \
    --bootstrap-server [***SRM SERVICE HOST***] \
    --config-file [***PROPERTIES FILE***] \
    --application-id srm-service

kafka-streams-application-reset \
    --bootstrap-server [***SRM SERVICE HOST***] \
    --config-file [***PROPERTIES FILE***] \
    --application-id srm-service-status
Replace [***PROPERTIES FILE***] with the location of a configuration file that contains all necessary security properties that are required to establish a connection with the Kafka service. This option is only required if your Kafka service is secured.

Streams Replication Manager – enable internal topics based remote topic discovery

The Remote Topics Discovery With Internal Topic property is not selected during the upgrade by default. Ensure that you select the property following an upgrade and restart the SRM service. This property enables SRM to track remote (replicated) topics using an internal Kafka topic instead of relying on the naming convention of the replication policy that is in use. This enables the SRM service to provide better monitoring insights on replications.

Streams Replication Manager – transition IdentityReplicationPolicy configuration

Cloudera Runtime 7.1.9 introduces a dedicated configuration property in Cloudera Manager to enable prefixless replication with the IdentityReplicationPolicy. If you have been using the IdentityReplicationPolicy in a previous version, Cloudera recommends that you transition your configuration to use the new configuration property. If you transition your configuration, you will be able to monitor your replications with the SRM Service, which was not possible in previous versions if the IdentityReplicationPolicy was in use.
  1. In Cloudera Manager, select the SRM service.
  2. Go to Configuration.
  3. Clear the following entry from Streams Replication Manager's Replication Config.
    replication.policy.class=org.apache.kafka.connect.mirror.IdentityReplicationPolicy
  4. Select Enable Prefixless Replication.
  5. Ensure that Remote Topics Discovery With Internal Topic is selected.

    If this property is not selected when the IdentityReplicationPolicy is in use, replication monitoring provided by the SRM Service does not function. This property is enabled by default when Enable Prefixless Replication is selected.

  6. Restart the SRM service.

Cruise Control – reconfigure goals to customized values

Provide the customized values for the Cruise Control goal sets, if you needed to copy them before upgrading your cluster. Furthermore, you must rename any mentioning of com.linkedin.kafka.cruisecontrol.analyzer.goals.RackAwareGoal to com.linkedin.kafka.cruisecontrol.analyzer.goals.RackAwareDistributionGoal as Cruise Control will not be able to start otherwise. For more information, see the Cruise Control Known Issues.

Cruise Control – configure Kerberos ATL rules

Cruise Control now uses the cluster-wide Kerberos auth-to-local (ATL) rules by default. A new configuration property called SPNEGO Principal to Local Rules is introduced. This property is used to manually specify the ATL rules. During an upgrade, the property is set to DEFAULT to ensure backward compatibility. If you want to use the cluster-wide Kerberos ATL rules for custom Kerberos principals, you need to clear the existing value from the SPNEGO Principal to Local Rules property in Cloudera Manager > Cruise Control > Configuration.

Migrating a YARN Queue Manager database to an external database

Post installation, you must migrate your YARN Queue Manager database to a PostgreSQL external database. This migration is needed because Queue Manager uses a different database as its backing store than what was used in previous versions of Cloudera Data Platform (CDP) 7.1.8 and below. Once you have configured the database as detailed in the steps below, the data migration process is automatic.
You must supply the details of a PostgreSQL database to migrate your old data to 7.1.9, as Queue Manager only uses PostgreSQL in CDP 7.1.9 or higher versions.
Before you begin: You must first create a database for Queue Manager and then configure the database to assign roles and permissions for the user (dbuser). The roles users need to assign are the following: SUPERUSER CREATEDB CREATEROLE INHERIT LOGIN. You can use the sample query or a similar query below:
CREATE ROLE qmadmin PASSWORD 'password' SUPERUSER CREATEDB CREATEROLE INHERIT LOGIN;
CREATE DATABASE "configstore";
ALTER DATABASE "configstore" OWNER TO qmadmin;
  1. In your Cloudera Manager instance, navigate to Clusters > Yarn Queue Manager.
  2. Click the Configuration tab.
  3. In the left navigation menu under Category, click Database.
  4. Enter the following fields that are required for the PostgreSQL database that is to be used. You will have these values from the query that was run in the prerequisite:
    • QueueManager Config Service Database Name
    • QueueManager Config Service Database Host
    • QueueManager Config Service Database Port
    • QueueManager Config Service Database User
    • QueueManager Config Service Database User Password
  5. Click Save Changes.
  6. In Queue Manager, navigate to Actions and click Restart. The data migration to the PostgreSQLdata database may take a few minutes after Queue Manager is restarted.