Cloudera Docs

Troubleshooting replication policies in CDP Public Cloud

The troubleshooting scenarios in this topic help you to troubleshoot issues in the Replication Manager service in CDP Public Cloud.

Different methods to identify errors related to failed replication policy

What are the different methods to identify errors while troubleshooting a failed replication policy?

You can choose one of the following methods to identify the errors to troubleshoot a job failure:
  • On the Replication Policies page, click the failed job in the Job History pane. The errors for the failed job appear.
    The following sample image shows the Job History pane for a replication policy job:
  • In the source and target Cloudera Manager, click Running Commands on the left navigation bar. The recent command history shows the failed commands.
    The following sample image shows the Running Commands page for an HBase replication policy:
  • On the source cluster and target cluster, open the service logs to track the errors (For example, HBase service logs).

    You can also search on the Cloudera Manager > Diagnostics > Logs page to view the logs.

HDFS replication policy fails due to export HTTPS_PROXY environment variable

HDFS replication policies fail when the export HTTPS_PROXY environment variable is set to access AWS through proxy servers. How to resolve this issue?

Remedy

To resolve this issue, perform the following steps:
  1. Open the core-site.xml file on the source cluster.
  2. Enter the following properties in the file: 
    <property>
  <name>fs.s3a.proxy.host</name>
  <description>Hostname of the (optional) proxy server for S3 connections.</description>
</property>

<property>
  <name>fs.s3a.proxy.port</name>
  <description>Proxy server port. If this property is not set
    but fs.s3a.proxy.host is, port 80 or 443 is assumed (consistent with
    the value of fs.s3a.connection.ssl.enabled).</description>
</property>
  3. Save and close the file.
  4. Restart the source Cloudera Manager.
  5. Run the failed HDFS replication policies in Replication Manager.
    Replication Manager completes the replication successfully.

Cannot find destination clusters for HBase replication policies

When you ping destination clusters using their host names, the source cluster hosts for HBase replication policies do not find the destination clusters. How to resolve this issue?

Cause

This might occur for on-premises clusters such as CDP Private Cloud Base clusters or CDH clusters because the source clusters are not on the same network as the destination Data Hub. Therefore, hostnames cannot be resolved by the DNS service on the source cluster.

Remedy

Add the destination Region Server and Zookeeper IP to host name mappings in the /etc/hosts files of all the Region Servers on the source cluster.
The following snippet shows the contents in a sample /etc/hosts file:
10.115.74.181 dx-7548-worker2.dx-hbas.x2-8y.dev.dr.work
10.115.72.28 dx-7548-worker1.dx-hbas.x2-8y.dev.dr.work
10.115.73.231 dx-7548-worker0.dx-hbas.x2-8y.dev.dr.work
10.115.72.20 dx-7548-master1.dx-hbas.x2-8y.dev.dr.work
10.115.74.156 dx-7548-master0.dx-hbas.x2-8y.dev.dr.work
10.115.72.70 dx-7548-leader0.dx-hbas.x2-8y.dev.dr.work

HBase replication policy fails when Perform Initial Snapshot is chosen

An HBase replication policy fails for COD on Microsoft Azure when the "Perform Initial Snapshot" option is chosen but data replication is successful when the option is not chosen. How to resolve this issue?

Cause

This issue appears when the required managed identity of source roles are not assigned.

Remedy

Assign the managed identity of source roles, Storage Blob Data Owner or Storage Blob Data Contributor, to the destination storage data container and vice versa for bidirectional replication.
The roles allow writing a snapshot in the destination cluster container.

Optimize HBase replication policy performance when replicating HBase tables with several TB data

Can HBase replication policy performance be optimized when replicating HBase tables with several TB of data if the "Perform Initial Snapshot" option is chosen during HBase replication policy creation?

Complete the following manual steps to optimize HBase replication policy performance when replicating several TB of HBase data if you choose the Perform Initial Snapshot option during the HBase replication policy creation process.

Remedy

  1. Before you create the HBase replication policy, perform the following steps:
    1. Navigate to the source Cloudera Manager > YARN service > Configuration tab.
    2. Search for the mapreduce.task.timeout parameter.
    3. Increase the value or set it to 0 to switch off the timeout.
    4. Restart the YARN service.
    5. Navigate to the source Cloudera Manager > HBase service > Configuration tab.
    6. Search and configure the following key-value pairs:
      • hbase.snapshot.master.timeout.millis = 840000

      • hbase.client.sync.wait.timeout.msec = 180000

      • hbase.client.operation.timeout = 2400000

      • hbase.client.procedure.future.get.timeout.msec = 3000000

      • hbase.hfilearchiver.thread.pool.max=100

      • hbase.snapshot.thread.pool.max=24

    7. Restart the HBase service.
    8. Perform steps e through g on the target Cloudera Manager.
  2. When you create the HBase replication policy for the first time using the above configured source cluster, you must increase the Maximum Map Slots value to a higher number on the Advanced Settings page.
  3. If Store File Tracking (SFT) is enabled in the target COD, perform the steps mentioned in the COD migration topic after the replication policy creation is complete.

Partition metadata replication takes a long time to complete

How can partition metadata replication be improved when the Hive tables use several Hive partitions?

Hive metadata replication process takes a long time to complete when the Hive tables use several Hive partitions. This is because the Hive partition parameters are compared during the import stage of the partition metadata replication process and if the exported and existing partition parameters do not match, the partition is dropped and recreated. You can configure a key-value pair to support partition metadata replication.

  1. Go to the Cloudera Manager > Clusters > Hive service > Configuration tab.
  2. Search for the Hive Replication Environment Advanced Configuration Snippet (Safety Valve) property.
  3. Enter the HIVE_IGNORED_PARTITION_PARAMETERS=[***comma separated list of Hive partition parameters***] key-value pair.
    For example, 
    HIVE_IGNORED_PARTITION_PARAMETERS=transient_lastDdlTime,totalSize,numRows,COLUMN_STATS_ACCURATE,numFiles

    The partition parameter names you provide are not compared during the import stage of the partition metadata replication process. Therefore, even if the partition parameters do not match between the exported and existing partitions, the partition is not dropped or recreated. After you configure this key-value pair, the import stage of the partition metadata replication process completes faster.

  4. Save the changes, and restart the Hive service.

Replicating Hive nested tables

CDP Public Cloud Replication Manager does not support Hive nested tables. What do I do if there are Hive nested tables in the source cluster?

CDP Public Cloud Replication Manager does not support Hive nested tables for replication. Therefore, it is recommended that you move the nested tables to a different location in HDFS and then replicate Hive external tables. However, if this is not possible, you can perform the following steps in the given order as a workaround.

Solution

  1. Create a Hive replication policy on the target cluster. Ensure that the Additional Settings > Replication Option > Metadata only option is selected to replicate the metadata of required files and directories.
  2. Create a HDFS replication policy on the source cluster to replicate the table data.

Target HBase folder is deleted when HBase replication policy fails

When the snapshot export fails during the HBase replication policy job run, the target HBase folder in the destination Data Hub or COD gets deleted.

You can either revoke the delete permission for the user, or ensure that you use an access key/role that does not have delete permissions to the required storage component.

The following steps show how to create an access key in AWS and an Azure service principal, which do not have delete permission for the storage component.

Solution in AWS

If the Data Hub or COD is on Amazon S3, you can perform the following steps to create an access key that does not have delete permission for the storage component:
  1. Login to AWS.
  2. Click Create policy on the IAM > Policies > Create policy page.
    The following sample image shows the Policies page in the AWS > IAM section to create a policy.
    The sample image shows the Policies page in IAM to create a policy in AWS.
  3. Choose the S3 Service, and then choose the following permissions for generic access (assign resources as required):
    • List/ListBucket
    • Read/GetBucketLocation
    • Read/GetObject
    • Write/AbortMultipartUpload
    • Write/PutObject
    • Permissions management/PutBucketPublicAccessBlock
  4. Add Delete/DeleteObject permission to the target COD cluster’s snapshot temporary folder.
    For example, the target COD cluster’s snapshot temporary folder might be located in [***target COD S3 path***]/hbase/.hbase-snapshot/.tmp/*.
  5. Enter a Name for the policy, add tags, and click Create policy.
  6. Click Add Users to create a user on the IAM > Users page.
  7. Enter a Name, and click Next.
  8. Choose the Attach policies directly option on the Set permission page, and then assign the previously created policy to the user.
  9. Optionally, add tags and create the user.
  10. Click Create access key to create an access key for the user on the IAM > Users > [***new user***] page.
  11. On the Security Credentials > Access keys page, choose Application running outside AWS. Click Next.
  12. Optionally, attach the tags, create and save the access key. This access key is used as an external account for replication.

    How do I verify whether the target HBase folder in the destination Data Hub or COD does not get deleted if the snapshot export fails during the HBase replication policy job run?

    Perform the following steps to verify if the delete operation is allowed for the access key that you previously created:
    1. Run the aws configure --profile delete-test command to setup the credentials in AWS CLI.
    2. Delete an arbitrarily created temporary file from the account using the aws s3 --profile delete-test rm --recursive s3://[***account name***]/delete-testing/ command.
    The delete operation is not allowed.

Solution in Microsoft Azure

If the Data Hub or COD is on ABFS, you can perform the following steps to create an Azure service principal that does not have delete permission for the storage component:
  1. Login to Microsoft Azure.
  2. Click Add > Add a custom role on the Subscriptions > Select the subscription > Access Control (IAM) page in Microsoft Azure, and complete the following steps:
    1. On the Basics tab, provide a name for the role, select Clone a role for the Baseline permissions field, and choose Storage Blob Data Contributor for the Role to clone field. Click Next.
    2. On the Permissions tab, remove the Microsoft.Storage/storageAccounts/blobServices/containers/delete and Microsoft.Storage/storageAccounts/blobServices/containers/blobs/delete permissions.
    3. Click Review + create.
      The following sample image shows the Access control (IAM) page in Microsoft Azure:
      The sample image shows the Access control (IAM) page in Microsoft Azure to create a custom role.
  3. Click Add > Add role assignment on the Storage accounts > Select the account where snapshots will be written > Access Control (IAM) page, and complete the following steps:
    1. On the Role tab, select the custom role previously created. Click Next.
    2. On the Members tab, select User, group, or service principal for Assign access to field, and select the required service principal.
    3. Click Review + assign.
    4. Click Review + assign on the Conditions (optional) tab.
  4. Click Add principal on the Storage accounts > [***account name***] > [***container where snapshots are to be written***] > Manage ACL page.
    1. Select the required service principal.
    2. Choose the Execute permission for the required container, and click Save.

    How do I verify whether the target HBase folder in the destination Data Hub or COD does not get deleted if the snapshot export fails during the HBase replication policy job run?

    To verify if the delete operation is allowed on the service principal that you previously created, perform the following steps:
    1. Open the Azure Cloud Shell terminal.
    2. Login using the service principal that you created previously using the az login --service-principal -u [***client id***] -p [***client secret***] --tenant [***tenant id***] command.
    3. Delete an arbitrarily created temporary file from the account using the az storage fs file delete --path [***temporary file***] -f data --account-name [***account name***] --auth-mode login command.
    The delete operation is not allowed.

Replicate HBase data in existing and future tables

Errors might appear when you try to replicate HBase data from existing tables and future tables in a database using the “Replicate Database” option during the HBase replication policy creation process. These errors appear when there are compatibility issues.

The following list shows a few errors that might appear and how to mitigate these issues:

  • The error Destination cluster doesn't support replication of all HBase tables. Please change the destination cluster or specify the tables that should be replicated. appears during HBase replication policy creation process. How to mitigate this issue?
    This error appears if you choose the Select Source > Replicate Database option during the HBase replication policy creation process for unsupported cluster versions.

    To mitigate this issue, ensure that the target Cloudera Manager version is 7.11.0 or higher and the source cluster version is CDH 6.x or higher before you choose the Select Source > Replicate Database option during the HBase replication policy creation process.

  • The error Exception in thread "main" java.lang.IllegalArgumentException: Cannot add a peer with id=_repl__f1907207cdbc4a528777bb4a316_ba23f09f7328494bbcbf81f40b because that id already exists appears after creating subsequent HBase replication policies.
    This error appears if you created an HBase peer manually using the hbase shell to replicate all column families on a source cluster using CDH 5.x.

    You can replicate the HBase data (existing tables and future tables) in a database only if the target Cloudera Manager version is 7.11.0 or higher and the source cluster version is CDH 6.x or higher. Therefore, even if you create a peer using hbase shell in the source cluster version CDH 5.16.2 or any other unsupported version, errors might appear.