[
https://issues.apache.org/jira/browse/HDFS-7716?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14306764#comment-14306764
]
Zhe Zhang commented on HDFS-7716:
---------------------------------
Thanks Jing! The high-level plan looks good. Below are some detailed items we
can further discuss:
# Shall we use {{triplets}} to record block locations in an block group?
# If we use {{triplets}} for both replicated blocks and EC block groups, I
don't clearly see which fields and methods to offload to
{{BlockReplicationInfo}}. Thoughts?
# We discussed recording the EC schema in {{BlockGroupInfo}} to avoid
frequently checking INode. Before HDFS-7337 we can simply record the number of
data and parity blocks (2 {{short}} variables).
# An INode feature for {{BlockGroupInfo}} sounds good. Before our meetup last
week, HDFS-7339's [patch |
https://issues.apache.org/jira/secure/attachment/12693731/HDFS-7339-006.patch]
used that approach. We might be able to reuse some code.
# [~andrew.wang] and I have discussed how to support snapshots when a file is
converted between replicated and EC forms. If we keep both the regular
{{BlockInfo}} array and the {{BlockGroupInfo}} feature, we should be able to
calculate snapshots.
> Erasure Coding: extend BlockInfo to handle EC info
> --------------------------------------------------
>
> Key: HDFS-7716
> URL: https://issues.apache.org/jira/browse/HDFS-7716
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: Jing Zhao
> Assignee: Jing Zhao
>
> The current BlockInfo's implementation only supports the replication
> mechanism. To use the same blocksMap handling block group and its data/parity
> blocks, we need to define a new BlockGroupInfo class.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
