Community Articles

Find and share helpful community-sourced technical articles.
Celebrating as our community reaches 100,000 members! Thank you!


Apache Atlas provides open metadata management and governance capabilities for organisations to build a catalog of their data assets, classify and govern these assets and provide collaboration capabilities around these data assets for data scientists, analysts and the data governance team.


This article focuses on backup and restore of Atlas data during HDP to CDP migration. 

Migration Paths 

HDP 2.6 to CDP 


Note: Following commands need to be executed on HDP Cluster

  • Download the migration script to download on to Atlas server host. 


## Download the script

## Untar
tar zxvf atlas-migration-exporter-

##copy the contents
mkdir /usr/hdp/
cp atlas-migration-exporter-* /usr/hdp/
chown -R atlas:hadoop /usr/hdp/


  • Before taking the backup Stop the atlas 


## Stop Atlas via ambari UI

## /root/atals_metadata is by backup directory this is an example

## Run the migration script
[root@ccycloud-1 ~]# python /usr/hdp/ -d /root/atlas_metadata
atlas-migration-export: starting migration export. Log file location /var/log/atlas/atlas-migration-exporter.log
atlas-migration-export: initializing
atlas-migration-export: ctor: parameters: 3
atlas-migration-export: initialized
atlas-migration-export: exporting typesDef to file /root/atlas_metadata/atlas-migration-typesdef.json
atlas-migration-export: exported  typesDef to file /root/atlas_metadata/atlas-migration-typesdef.json
atlas-migration-export: exporting data to file /root/atlas_metadata/atlas-migration-data.json
atlas-migration-export: exported  data to file /root/atlas_metadata/atlas-migration-data.json
atlas-migration-export: completed migration export!


  • Make sure the exported data files in json format found in the location 


## make sure the backup is available 
[root@ccycloud-1 ~]# ls -ltrh /root/atlas_metadata
total 240K
-rw-r--r-- 1 root root  32K Sep  8 02:57 atlas-migration-typesdef.json
-rw-r--r-- 1 root root 205K Sep  8 02:57 atlas-migration-data.json




Note: Following commands need to be executed on CDP Cluster


Note: For this migration Atlas must be empty before following the next steps.  

To restore atlas metadata we need to start Atlas in migration mode this can be done by 

* configure via CM UI >> Atlas >>conf/atlas-application.properties_role_safety_valve

Screenshot 2022-07-15 at 11.38.42.png

  • Start the Atlas and wait until the migration status is completed

Screenshot 2022-07-15 at 11.38.48.png

Screenshot 2022-07-15 at 11.38.57.png

  • After the migration is completed as shown above. Please remove the config and restart atlas again. Expect to see a status as shown below. 

Screenshot 2022-07-15 at 11.39.12.png

HDP 3.X to CDP 


Note: Following commands need to be executed on HDP Cluster

Atlas metadata is stored at Hbase and Infra-solr collections. Both locations need to be backed up. 

  • Hbase backup 


### script this
hbase shell 

disable 'atlas_janus'
snapshot 'atlas_janus', 'atlas_janus-backup-new'
enable 'atlas_janus'

## Linux cli
hbase org.apache.hadoop.hbase.snapshot.ExportSnapshot -snapshot 'atlas_janus-backup-new' -copy-to hdfs:///tmp/hbase_new_atlas_backups
hbase org.apache.hadoop.hbase.snapshot.ExportSnapshot -snapshot 'ATLAS_ENTITY_AUDIT_EVENTS-backup-new' -copy-to hdfs:///tmp/hbase_new_atlas_backups



  • Infra-solr collections backup


## Getting Kerberos ticket 
klist -kt /etc/security/keytabs/ambari-infra-solr.service.keytab infra-solr/`hostname -f`@REALM

### Taking Dump of atlas collections e.g vertex_index (speed 1min/million records)

/usr/lib/ambari-infra-solr-client/ --zookeeper-connect-string zookeeper_host:2181/infra-solr --jaas-file /etc/ambari-infra-solr/conf/infra_solr_jaas.conf --dump-documents --collection vertex_index --output /home/solr/backup/atlas/vertex_index/data --max-read-block-size 100000 --max-write-block-size 100000 

### Replace the collection and back up edge_index,fulltext_index,vertex_index on at a time. 




Note: Following commands need to be executed on CDP Cluster

Use the backup of Hbase and Infra-solr collections restore them in to CDP environment before starting atlas in CDP world. 

  • Hbase restore tables from snapshot


##copy the backup directories to Target HDFS and run the restore commands 
hbase shell 

restore_snapshot 'atlas_janus-backup-new'

restore_snapshot 'ATLAS_ENTITY_AUDIT_EVENTS-backup-new' 


  • Infra-solr restore


## Getting Kerberos ticket 
klist -kt /etc/security/keytabs/ambari-infra-solr.service.keytab infra-solr/`hostname -f`@REALM

### Restoring your collections e.g vertex_index (speed 1min/million records)

/usr/lib/ambari-infra-solr-client/ --zookeeper-connect-string zookeeper_host:2181/solr-infra --jaas-file /etc/ambari-infra-solr/conf/infra_solr_jaas.conf --upload-documents --collection vertex_index --output /home/solr/backup/atlas/vertex_index/data --max-read-block-size 100000 --max-write-block-size 100000 

### make sure to restore all 3 collections i.e vertex_index, fulltext_index, edge_index


  • After restoring the Hbase tables and Solr collections. Start the Atlas via CM UI and validate the metrics. 

Useful Links & Scripts

  •  Command to download the metrics of atlas data. This metrics can be used for validation before and after migration. 


# curl -g -X GET -u admin:admin -H "Content-Type: application/json"  -H"Cache-Control: no-cache" "" >atlas_metrics.json
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100   261    0   261    0     0     64      0 --:--:--  0:00:04 --:--:--    64

# ls -ltrh atlas_metrics.json
-rw-r--r-- 1 root root 261 Sep  8 05:47 atlas_metrics.json


  • Atlas also comes with API to export and import entities sample code can be found here 

  • Known issue during the migration is the namespace of Hbase tables, make sure the tables are restored in default namespace if not configure the right namespace in atlas configs as below 

Screenshot 2022-07-15 at 13.18.08.png

Full Disclosure & Disclaimer:

I am an Employee of Cloudera, but this is not part of the formal documentation of the Cloudera Data platform. It is purely based on my own experience of advising people in their choice of tooling and customisations.

0 Kudos