Changing Ranger audit storage location and migrating data

How to change the location of existing and future Ranger audit data collected by Solr from HDFS to a local file system or from a local file system to HDFS.

Stop Atlas from Cloudera Manager.

If using Kerberos, set the SOLR_PROCESS_DIR environment variable.

# export SOLR_PROCESS_DIR=$(ls -1dtr /var/run/cloudera-scm-agent/process/*SOLR_SERVER | tail -1)

Starting with Cloudera Runtine version 7.1.4 / 7.2.2, the storage location for ranger audit data collected by Solr changed to local file system from HDFS, as was true for previous versions. The default storage location Ranger audit data storage location for Cloudera Runtine-7.1.4+ and Cloudera Runtine-7.2.2+ installations is local file system. After upgrading from an earlier Cloudera platform version, follow these steps to backup and migrate your Ranger audit data and change the location where Solr stores your future Ranger audit records.

The default value of the index storage in the local file system is /var/lib/solr-infra. You can configure this, using Cloudera Manager > Solr > Configuration > "Solr Data Directory".
The default value of the index storage in HDFS is /solr-infra. You can configure this, using Cloudera Manager > Solr > Configuration > "HDFS Data Directory".

Create HDFS Directory to store the collection backups.

As an HDFS super user, run the following commands to create the backup directory:
```
# hdfs dfs -mkdir /solr-backups
# hdfs dfs -chown solr:solr /solr-backups
```

Obtain valid kerberos ticket for Solr user.

# kinit -kt solr.keytab solr/$(hostname -f)

Download the configs for the collection.

# solrctl instancedir --get ranger_audits /tmp/ranger_audits
# solrctl instancedir --get atlas_configs /tmp/atlas_configs

Modify the solrconfig.xml for each of the configs for which data needs to be stored in HDFS.

In /tmp/<config_name>/conf created during Step 3., edit properties in the solrconfig.xml file as follows:

When migrating your data storage location from a local file system to HDFS, replace these two lines:

<directoryFactory name="DirectoryFactory" class="${solr.directoryFactory:solr.NRTCachingDirectoryFactory}">
<lockType>${solr.lock.type:native}</lockType>

with

<directoryFactory name="DirectoryFactory" class="${solr.directoryFactory:org.apache.solr.core.HdfsDirectoryFactory}">
<lockType>${solr.lock.type:hdfs}</lockType>

When migrating your data storage location from HDFS to a local file system, replace these two lines:

<directoryFactory name="DirectoryFactory" class="${solr.directoryFactory:org.apache.solr.core.HdfsDirectoryFactory}">
<lockType>${solr.lock.type:hdfs}</lockType>

with

<directoryFactory name="DirectoryFactory" class="${solr.directoryFactory:solr.NRTCachingDirectoryFactory}">
<lockType>${solr.lock.type:native}</lockType>

Backup the Solr collections.
- When migrating your data storage location from a local file system to HDFS, run:
```
# curl -k --negotiate -u : "https://$(hostname
 -f):8995/solr/admin/collections?action=BACKUP&name=vertex_backup&collection=vertex_index&
location=hdfs://<Namenode_Hostname>:8020/solr-backups"
```
  In the preceding command, the important points are name, collection, and location:
  
  name
  
  specifies the name of the backup. It should be unique per collection
  
  collection
  
  specifies the collection name for which the backup will be performed
  
  location
  
  specifies the HDFS path, where the backup will be stored
  
  Repeat the curl command for different collections, modifying the parameters as necessary for each collection.
  The expected output would be -
```
"responseHeader":{
 "status":0,
 "QTime":10567},
"success":{
 "Solr_Server_Hostname:8995_solr":{
   "responseHeader":{
   "status":0,
   "QTime":8959}}}}
```
- When migrating your data storage location from HDFS to a local file system:
  Refer to Back up a Solr collection for specific steps, and make the following adjustments:
  - If TLS is enabled for the Solr service, specify the trust store and password by using the ZKCLI_JVM_FLAGS environment variable before you begin the procedure.
```
# export ZKCLI_JVM_FLAGS="-Djavax.net.ssl.trustStore=/path/to/truststore.jks -Djavax.net.ssl.trustStorePassword="
```
  - Create Snapshot
```
# solrctl --jaas $SOLR_PROCESS_DIR/jaas.conf collection --create-snapshot <snapshot_name> -c <collection_name>
```
  - or use the Solr API to take the backup:
```
curl -i -k --negotiate -u : "https://(hostname -f):8995/solr/admin/collections?action=BACKUP&name=ranger_audits_bkp&collection=ranger_audits&location=/path/to/solr-backups"
```
  - Export Snapshot
```
# solrctl --jaas $SOLR_PROCESS_DIR/jaas.conf collection --export-snapshot <snapshot_name> -c <collection_name> -d <destination_directory>
```
    note
    The <destination_directory> is a HDFS path. The ownership of this directory should be solr:solr.

Update the modified configs in Zookeeper.

# solrctl --jaas $SOLR_PROCESS_DIR/jaas.conf instancedir --update atlas_configs /tmp/atlas_configs
# solrctl --jaas $SOLR_PROCESS_DIR/jaas.conf instancedir --update ranger_audits /tmp/ranger_audits

Delete the collections from the original location.

All instances of Solr service should be up, running, and healthy before deleting the collections. Use Cloudera Manager to check for any alerts or warnings for any of the instances. If alerts or warnings exist, fix those before deleting the collection.
```
# solrctl collection --delete edge_index
# solrctl collection --delete vertex_index
# solrctl collection --delete fulltext_index
# solrctl collection --delete ranger_audits
```
Verify that the collections are deleted from the original location.
```
# solrctl collection --list
```
This will give an empty result.
Verify that no leftover directories for any of the collections have been deleted.
- When migrating your data storage location from a local file system to HDFS:
```
# cd /var/lib/solr-infra
```
  Get the value of "Solr Data Directory, using Cloudera Manager > Solr > Configuration.
```
# ls -ltr
```
- When migrating your data storage location from HDFS to a local file system, replace these two lines:
```
# hdfs dfs -ls /solr/<collection_name>
```
  note
  If any directory name which starts with the collection name deleted in Step 7. exists, delete/move the directory to another path.
Restore the collection from backup to the new location.

Refer to Restore a Solr collection, for more specific steps.
```
# curl -k --negotiate -u : "https://$(hostname
 -f):8995/solr/admin/collections?action=RESTORE&name=<Name_of_backup>&location=hdfs:/
<<Namenode_Hostname>:8020/solr-backups&collection=<Collection_Name>"
```
```
# solrctl collection --restore ranger_audits
 -l hdfs://<Namenode_Hostname>:8020/solr-backups
 -b ranger_backup -i ranger1
```
The request id must be unique for each restore operation, as well as for each retry.
To check the status of restore operation:
```
# solrctl collection --request-status <requestId>
```
note
If the Atlas Collections (vertex_index, fulltext_index and edge_index) restore operations fail, restart the solr service and rerun the restore command. Now, the restart operations should complete successfully.
Verify the Atlas & Ranger functionality.
Verify that both Atlas and Ranger audits functions properly, and that you can see the latest audits in Ranger Web UI and latest lineage in Atlas.
- To verify Atlas audits, create a test table in Hive, and then query the collections to see if you are able to view the data.
- You can also query the collections every 20-30 seconds (depending on how other services utilize Atlas/Ranger), and verify if the "numDocs" value increases at every query.
```
# curl -k --negotiate -u : "https://$(hostname -f):8995/solr/edge_index/select?q=*%3A*&wt=json&ident=true&rows=0"
# curl -k --negotiate -u : "https://$(hostname -f):8995/solr/vertex_index/select?q=*%3A*&wt=json&ident=true&rows=0"
# curl -k --negotiate -u : "https://$(hostname -f):8995/solr/fulltext_index/select?q=*%3A*&wt=json&ident=true&rows=0"
# curl -k --negotiate -u : "https://$(hostname -f):8995/solr/ranger_audits/select?q=*%3A*&wt=json&ident=true&rows=0"
```