[
https://issues.apache.org/jira/browse/HDFS-4642?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13623544#comment-13623544
]
Harsh J commented on HDFS-4642:
-------------------------------
The proposed API would appear highly odd compared to the rest of the API design
right now. What your description seems to suggest is merely a simple bulk
alternative to the existing lease-recovery call which can take several paths to
work on under a single RPC call. How much of an advantage are you expecting to
find here versus issuing them in parallel from the dfsclient itself?
Lets keep in mind that to process a bulk of files, the time required increases
at the per-call level as well (similar problem to how bulkload requests in
HBase get retried just cause they didn't respond back in time), which is why
I'd also be hesitant at adding this unless there's a great deal to gain (as
opposed to clients issuing it in parallel).
> Allow lease recovery for multiple paths to be issued in one request
> -------------------------------------------------------------------
>
> Key: HDFS-4642
> URL: https://issues.apache.org/jira/browse/HDFS-4642
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: Ted Yu
>
> Currently client can only request lease recovery for one Path:
> {code}
> public boolean recoverLease(Path f) throws IOException {
> {code}
> For HBase distributed log splitting, Nicolas made a suggestion here:
> https://issues.apache.org/jira/browse/HBASE-7878?focusedCommentId=13615364&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13615364
> HBase master collects the files that should be split, it issues lease
> recovery for the files (in one request), then distribute log splitting.
> This would help shorten MTTR.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
