[
https://issues.apache.org/jira/browse/HDFS-7896?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14349792#comment-14349792
]
Arpit Agarwal commented on HDFS-7896:
-------------------------------------
Based on offline discussion with [~sanjay.radia].
Couple of options:
# Track IO latencies per volume to detect outliers. Unfortunately DNs use
buffered IO so IO latency is a meaningless for measuring disk performance.
# Run a microbenchmark on DN startup and tag slow disks. Can affect DN startup
performance and does not address disks that develop faults once the DN has been
started.
[~cnauroth] recently added failed volume reporting via HDFS-7604. Ideally we
can extend that reporting infrastructure.
> HDFS Slow disk detection
> ------------------------
>
> Key: HDFS-7896
> URL: https://issues.apache.org/jira/browse/HDFS-7896
> Project: Hadoop HDFS
> Issue Type: New Feature
> Components: datanode
> Reporter: Arpit Agarwal
>
> HDFS should detect slow disks. To start with we can flag this information via
> the NameNode web UI. Alternatively DNs can avoid using slow disks for writes.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
