hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Suresh Srinivas (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HDFS-1359) HDFS federation: Add BlockPoolID to block
Date Fri, 27 Aug 2010 17:53:58 GMT

    [ https://issues.apache.org/jira/browse/HDFS-1359?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12903495#action_12903495

Suresh Srinivas commented on HDFS-1359:

Some relevant sections from the design document attached to HDFS-1052. For more details see
the design document:
#	A Block Pool is an independent set of blocks that belongs to a single entity (e.g. namespace).
A block pool is managed independently of other pools, allowing owning entities to generate
Block IDs for new blocks without the need for co-ordination with other entities.
# Block Pool ID: A block pool ID identifies a block pool. It is a unique ID across all clusters.
When a new namespace is created (as part of the format operation of a NN) a unique id is generated
and stored persistently by the namenodes. Using a unique BlockPoolID simplifies the creation,
as it is done on the fly on the namenodes instead of an error prone alternative of an administrator
configuring it. Namenodes persist the BlockPoolID information on the disk and reuse it in
subsequent restarts.
# Why is BlockPoolID globally unique?
#* Rather than assure the uniqueness of block pool ID via admin configuration, a unique id
can be generated programmatically.
#* Clearly it has to be unique within a cluster. It is equally easy to make them globally
unique or cluster unique. Only cost is # of bits.
#* If BPID is not unique, due to the possibility of reusing a deleted BPID, two instances
of block pools might conflict.  An associated birthmark is needed to distinguish the two instances.
#* Allows cluster merges, where block pools from the clusters can be retained.

> HDFS federation: Add BlockPoolID to block
> -----------------------------------------
>                 Key: HDFS-1359
>                 URL: https://issues.apache.org/jira/browse/HDFS-1359
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>            Reporter: Suresh Srinivas
> This is a subtask to HDFS-1052 to add BlockPoolID to Block.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message