[
https://issues.apache.org/jira/browse/HDFS-5182?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13891042#comment-13891042
]
Haohui Mai commented on HDFS-5182:
----------------------------------
It looks like to me that passing the file descriptor is a cleaner design:
# If the data is shared through mmap(), an I/O error will trigger a SIGBUS on
the client side.
# It requires taking up the virtual memory space for the client, it could be
problematic if the client is running in a container.
# From a portability prospective, {{DuplicateHandle}} on Windows can easily
pass file descriptors around.
I might have missed something. [~cmccabe], can you please elaborate why you
prefer choice #1 instead of #2? Thanks.
> BlockReaderLocal must allow zero-copy reads only when the DN believes it's
> valid
> ---------------------------------------------------------------------------------
>
> Key: HDFS-5182
> URL: https://issues.apache.org/jira/browse/HDFS-5182
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: hdfs-client
> Affects Versions: 3.0.0
> Reporter: Colin Patrick McCabe
> Assignee: Colin Patrick McCabe
>
> BlockReaderLocal must allow zero-copy reads only when the DN believes it's
> valid. This implies adding a new field to the response to
> REQUEST_SHORT_CIRCUIT_FDS. We also need some kind of heartbeat from the
> client to the DN, so that the DN can inform the client when the mapped region
> is no longer locked into memory.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
