HDFS

This engine provides integration with the Apache Hadoop ecosystem by allowing to manage data on HDFS via ClickHouse. This engine is similar to the File and URL engines, but provides Hadoop-specific features.

This feature is not supported by ClickHouse engineers, and it is known to have a sketchy quality. In case of any problems, fix them yourself and submit a pull request.

Usage

ENGINE = HDFS(URI, format)

Engine Parameters

• URI - whole file URI in HDFS. The path part of URI may contain globs. In this case the table would be readonly.
• format - specifies one of the available file formats. To perform SELECT queries, the format must be supported for input, and to perform INSERT queries – for output. The available formats are listed in the Formats section.
• [PARTITION BY expr]

PARTITION BY

PARTITION BY — Optional. In most cases you don't need a partition key, and if it is needed you generally don't need a partition key more granular than by month. Partitioning does not speed up queries (in contrast to the ORDER BY expression). You should never use too granular partitioning. Don't partition your data by client identifiers or names (instead, make client identifier or name the first column in the ORDER BY expression).

For partitioning by month, use the toYYYYMM(date_column) expression, where date_column is a column with a date of the type Date. The partition names here have the "YYYYMM" format.

Example:

1. Set up the hdfs_engine_table table:

CREATE TABLE hdfs_engine_table (name String, value UInt32) ENGINE=HDFS('hdfs://hdfs1:9000/other_storage', 'TSV')

2. Fill file:

INSERT INTO hdfs_engine_table VALUES ('one', 1), ('two', 2), ('three', 3)

3. Query the data:

SELECT * FROM hdfs_engine_table LIMIT 2

┌─name─┬─value─┐
│ one  │     1 │
│ two  │     2 │
└──────┴───────┘

Implementation Details

• Reads and writes can be parallel.

• Not supported:

  • ALTER and SELECT...SAMPLE operations.
  • Indexes.
  • Zero-copy replication is possible, but not recommended.
  Zero-copy replication is not ready for production

  Zero-copy replication is disabled by default in ClickHouse version 22.8 and higher. This feature is not recommended for production use.

Globs in path

Multiple path components can have globs. For being processed file should exists and matches to the whole path pattern. Listing of files determines during SELECT (not at CREATE moment).

• * — Substitutes any number of any characters except / including empty string.
• ? — Substitutes any single character.
• {some_string,another_string,yet_another_one} — Substitutes any of strings 'some_string', 'another_string', 'yet_another_one'.
• {N..M} — Substitutes any number in range from N to M including both borders.

Constructions with {} are similar to the remote table function.

Example

1. Suppose we have several files in TSV format with the following URIs on HDFS:

   • 'hdfs://hdfs1:9000/some_dir/some_file_1'
   • 'hdfs://hdfs1:9000/some_dir/some_file_2'
   • 'hdfs://hdfs1:9000/some_dir/some_file_3'
   • 'hdfs://hdfs1:9000/another_dir/some_file_1'
   • 'hdfs://hdfs1:9000/another_dir/some_file_2'
   • 'hdfs://hdfs1:9000/another_dir/some_file_3'

2. There are several ways to make a table consisting of all six files:

CREATE TABLE table_with_range (name String, value UInt32) ENGINE = HDFS('hdfs://hdfs1:9000/{some,another}_dir/some_file_{1..3}', 'TSV')

Another way:

CREATE TABLE table_with_question_mark (name String, value UInt32) ENGINE = HDFS('hdfs://hdfs1:9000/{some,another}_dir/some_file_?', 'TSV')

Table consists of all the files in both directories (all files should satisfy format and schema described in query):

CREATE TABLE table_with_asterisk (name String, value UInt32) ENGINE = HDFS('hdfs://hdfs1:9000/{some,another}_dir/*', 'TSV')

Note

If the listing of files contains number ranges with leading zeros, use the construction with braces for each digit separately or use ?.

Example

Create table with files named file000, file001, ... , file999:

CREATE TABLE big_table (name String, value UInt32) ENGINE = HDFS('hdfs://hdfs1:9000/big_dir/file{0..9}{0..9}{0..9}', 'CSV')

Configuration

Similar to GraphiteMergeTree, the HDFS engine supports extended configuration using the ClickHouse config file. There are two configuration keys that you can use: global (hdfs) and user-level (hdfs_*). The global configuration is applied first, and then the user-level configuration is applied (if it exists).

  <!-- Global configuration options for HDFS engine type -->
  <hdfs>
    <hadoop_kerberos_keytab>/tmp/keytab/clickhouse.keytab</hadoop_kerberos_keytab>
    <hadoop_kerberos_principal>clickuser@TEST.CLICKHOUSE.TECH</hadoop_kerberos_principal>
    <hadoop_security_authentication>kerberos</hadoop_security_authentication>
  </hdfs>

  <!-- Configuration specific for user "root" -->
  <hdfs_root>
    <hadoop_kerberos_principal>root@TEST.CLICKHOUSE.TECH</hadoop_kerberos_principal>
  </hdfs_root>

Configuration Options

Supported by libhdfs3

parameter	default value
rpc_client_connect_tcpnodelay	true
dfs_client_read_shortcircuit	true
output_replace-datanode-on-failure	true
input_notretry-another-node	false
input_localread_mappedfile	true
dfs_client_use_legacy_blockreader_local	false
rpc_client_ping_interval	10 * 1000
rpc_client_connect_timeout	600 * 1000
rpc_client_read_timeout	3600 * 1000
rpc_client_write_timeout	3600 * 1000
rpc_client_socket_linger_timeout	-1
rpc_client_connect_retry	10
rpc_client_timeout	3600 * 1000
dfs_default_replica	3
input_connect_timeout	600 * 1000
input_read_timeout	3600 * 1000
input_write_timeout	3600 * 1000
input_localread_default_buffersize	1 1024 1024
dfs_prefetchsize	10
input_read_getblockinfo_retry	3
input_localread_blockinfo_cachesize	1000
input_read_max_retry	60
output_default_chunksize	512
output_default_packetsize	64 * 1024
output_default_write_retry	10
output_connect_timeout	600 * 1000
output_read_timeout	3600 * 1000
output_write_timeout	3600 * 1000
output_close_timeout	3600 * 1000
output_packetpool_size	1024
output_heartbeat_interval	10 * 1000
dfs_client_failover_max_attempts	15
dfs_client_read_shortcircuit_streams_cache_size	256
dfs_client_socketcache_expiryMsec	3000
dfs_client_socketcache_capacity	16
dfs_default_blocksize	64 1024 1024
dfs_default_uri	"hdfs://localhost:9000"
hadoop_security_authentication	"simple"
hadoop_security_kerberos_ticket_cache_path	""
dfs_client_log_severity	"INFO"
dfs_domain_socket_path	""

HDFS Configuration Reference might explain some parameters.

ClickHouse extras

parameter	default value
hadoop_kerberos_keytab	""
hadoop_kerberos_principal	""
libhdfs3_conf	""

Limitations

• hadoop_security_kerberos_ticket_cache_path and libhdfs3_conf can be global only, not user specific

Kerberos support

If the hadoop_security_authentication parameter has the value kerberos, ClickHouse authenticates via Kerberos. Parameters are here and hadoop_security_kerberos_ticket_cache_path may be of help. Note that due to libhdfs3 limitations only old-fashioned approach is supported, datanode communications are not secured by SASL (HADOOP_SECURE_DN_USER is a reliable indicator of such security approach). Use tests/integration/test_storage_kerberized_hdfs/hdfs_configs/bootstrap.sh for reference.

If hadoop_kerberos_keytab, hadoop_kerberos_principal or hadoop_security_kerberos_ticket_cache_path are specified, Kerberos authentication will be used. hadoop_kerberos_keytab and hadoop_kerberos_principal are mandatory in this case.

HDFS Namenode HA support

libhdfs3 support HDFS namenode HA.

• Copy hdfs-site.xml from an HDFS node to /etc/clickhouse-server/.
• Add following piece to ClickHouse config file:

  <hdfs>
    <libhdfs3_conf>/etc/clickhouse-server/hdfs-site.xml</libhdfs3_conf>
  </hdfs>

• Then use dfs.nameservices tag value of hdfs-site.xml as the namenode address in the HDFS URI. For example, replace hdfs://appadmin@192.168.101.11:8020/abc/ with hdfs://appadmin@my_nameservice/abc/.

Virtual Columns

• _path — Path to the file. Type: LowCardinalty(String).
• _file — Name of the file. Type: LowCardinalty(String).
• _size — Size of the file in bytes. Type: Nullable(UInt64). If the size is unknown, the value is NULL.
• _time — Last modified time of the file. Type: Nullable(DateTime). If the time is unknown, the value is NULL.

Storage Settings

• hdfs_truncate_on_insert - allows to truncate file before insert into it. Disabled by default.
• hdfs_create_new_file_on_insert - allows to create a new file on each insert if format has suffix. Disabled by default.
• hdfs_skip_empty_files - allows to skip empty files while reading. Disabled by default.

See Also

• Virtual columns