[
https://issues.apache.org/jira/browse/HDFS-6465?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14021786#comment-14021786
]
Dian Fu commented on HDFS-6465:
-------------------------------
update some design details about configurations:
requirements:
1. Existing deployments must be able to use the existing configuration
without any change.
2. As many as possible the configurations for different clusters must be
the same. The special configuration required for different clusters should be
minimal.
Configurations added:
• DFS_REGION_ID("dfs.region.id") : the region id of current cluster
• DFS_REGIONS("dfs.regions") : the region ids of all clusters, including
both the primary cluster and mirror clusters
• DFS_REGION_PRIMARY("dfs.region.primary") : the region id of primary
cluster
Configurations must be suffixed with regionId:
DFS_NAMENODE_RPC_ADDRESS_KEY, DFS_NAMENODE_SERVICE_RPC_ADDRESS_KEY,
DFS_NAMENODE_HTTP_ADDRESS_KEY, DFS_NAMENODE_HTTPS_ADDRESS_KEY,
DFS_NAMENODE_SECONDARY_HTTP_ADDRESS_KEY and DFS_NAMENODE_BACKUP_ADDRESS_KEY
Configurations could be suffixed with regionId or not.
These include all the configurations in NameNode.NAMENODE_SPECIFIC_KEYS and
NameNode.NAMESERVICE_SPECIFIC_KEYS except the above configurations which must
be suffixed with regionId:
DFS_NAMENODE_RPC_BIND_HOST_KEY, DFS_NAMENODE_NAME_DIR_KEY,
DFS_NAMENODE_EDITS_DIR_KEY, DFS_NAMENODE_SHARED_EDITS_DIR_KEY,
DFS_NAMENODE_CHECKPOINT_DIR_KEY, DFS_NAMENODE_CHECKPOINT_EDITS_DIR_KEY,
DFS_NAMENODE_SERVICE_RPC_BIND_HOST_KEY, DFS_NAMENODE_HTTP_BIND_HOST_KEY,
DFS_NAMENODE_HTTPS_BIND_HOST_KEY, DFS_NAMENODE_KEYTAB_FILE_KEY,
DFS_NAMENODE_SECONDARY_HTTPS_ADDRESS_KEY,
DFS_SECONDARY_NAMENODE_KEYTAB_FILE_KEY, DFS_NAMENODE_BACKUP_HTTP_ADDRESS_KEY,
DFS_NAMENODE_BACKUP_SERVICE_RPC_ADDRESS_KEY,
DFS_NAMENODE_KERBEROS_PRINCIPAL_KEY,
DFS_NAMENODE_KERBEROS_INTERNAL_SPNEGO_PRINCIPAL_KEY, DFS_HA_FENCE_METHODS_KEY,
DFS_HA_ZKFC_PORT_KEY and DFS_HA_AUTO_FAILOVER_ENABLED_KEY
The above configurations can be configured in the following format to
distinguish between clusters:
<configuration key>.<nameservice id>.<namenode id>.<region id>
If a configuration with a region id as suffix cannot be found, the
configuration without region id as suffix will be used instead.
All other configurations which aren’t mentioned should not be suffixed with
regionId.
> Enable the configuration of multiple clusters
> ---------------------------------------------
>
> Key: HDFS-6465
> URL: https://issues.apache.org/jira/browse/HDFS-6465
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: Dian Fu
> Assignee: Dian Fu
> Attachments: HDFS-6465.1.patch, HDFS-6465.2.patch, HDFS-6465.patch
>
>
> Tracks the changes required for configuration DR.
> configurations added:
> DFS_REGION_ID("dfs.region.id") : the region id of current cluster
> DFS_REGIONS("dfs.regions") : the region ids of all clusters, including
> both the primary cluster and mirror cluster
> DFS_REGION_PRIMARY("dfs.region.primary") : the region id of primary
> cluster
> configurations modified:
> The configurations in NAMENODE.NAMENODE_SPECIFIC_KEYS can be configured
> in the following format to distinguish between clusters.
> If a configuration with a region id as suffix cannot be found, the
> configuration without region id as suffix will be used instead:
> <configuration key>.<nameservice id>.<namenode id>.<region id>
> The configurations in NAMENODE.NAMESERVICE_SPECIFIC_KEYS can be configured in
> the following format to distinguish between clusters.
> If a configuration with a region id as suffix cannot be found, the
> configuration without region id as suffix will be used instead:
> <configuration key>.<nameservice id>.<region id>
--
This message was sent by Atlassian JIRA
(v6.2#6252)
