Download and Ingest

Nebula Graph uses RocksDB as the default key-value storage engine. Therefore, when you are trying to load a large amount of data, the SST files of RocksDB can be generated offline by running a map-reduce job and distributed directly to the servers.

Nebula Graph provides Spark-SSTFile-Generator tool.

Spark-SSTFile-Generator generates SST files from the hive table via the mapping files. For details on how to use it, please refer Spark application command line reference.

The execution will generate SST files on HDFS. The directory structure is as follows:

|---1 (this is partition number)
|        | ---- vertex-${FIRST_KEY_IN_THIS_FILE}.sst
|        | ---- edge-${FIRST_KEY_IN_THIS_FILE}.sst
         | ---- vertex-${FIRST_KEY_IN_THIS_FILE}.sst
         | ---- edge-${FIRST_KEY_IN_THIS_FILE}.sst

Each directory is a partition number.

SST file name format is {TYPE}-${FIRST_KEY_IN_THIS_FILE}.sst, where TYPE is data type, FIRST_KEY_IN_THIS_FILE is the start Key of the file. (If you want to write your own tools to generate SST files, you need to ensure that the keys in each SST file are ordered.)

Please confirm that all servers have Hadoop installed and HADOOP_HOME set.

Run Nebula Graph console and execute the download command:


Download the SST files of each server into directory data/download respectively via the DOWNLOAD command and meta in the storage servers. Explanation of the above command:

  • HADOOP_HOST specifies Hadoop NameNode address
  • HADOOP_PORT specifies Hadoop NameNode port number
  • HADOOP_PATH specifies Hadoop data storage directory

If error occurs when downloading, delete the corresponding data files in data/download directory and try to download again. If error occurs again, please raise us an issue on GitHub. When data download is done, re-execute the command leads to no actions.

When the offline SST data download is done, it can be ingested into the storage service via INGEST command. INGEST command is as follows:

nebula > INGEST

The command will ingest the SST files in data/download directory.

Note: ingest will block RocksDB when the data amount is large, please avoid running the command at requirement peak.