[
https://issues.apache.org/jira/browse/HDFS-6982?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14211634#comment-14211634
]
Andrew Wang commented on HDFS-6982:
-----------------------------------
I triggered another run manually, haven't looked at latest patch yet, but
responding to your above:
bq. The idea was to increase the periods in exponential manner: 5^0,5^1, 5^2
Sure, if you prefer :) It's all configurable, so I'm easy on this one.
bq. I am inclined towards redundantly specifying the audit logger in the conf.
I think it would also avoid confusion for future readers if we spell out the
registered audit loggers.
I see nntop as a new set of metrics, and the fact that it hooks into the audit
logger as an implementation detail that doesn't need to be exposed to clients.
Are there any downsides to automatically registering the TopAuditLogger besides
conf visibility? If there isn't already, we can add a print at NN startup that
lists all the audit loggers.
Generally speaking, simplifying the conf is something we try really hard to do.
If we do decide to redundantly specify, we'll need to beef up the docs in
hdfs-default.xml a bit to explain this.
> nntop: top-like tool for name node users
> -----------------------------------------
>
> Key: HDFS-6982
> URL: https://issues.apache.org/jira/browse/HDFS-6982
> Project: Hadoop HDFS
> Issue Type: New Feature
> Reporter: Maysam Yabandeh
> Assignee: Maysam Yabandeh
> Attachments: HDFS-6982.patch, HDFS-6982.v2.patch, HDFS-6982.v3.patch,
> HDFS-6982.v4.patch, HDFS-6982.v5.patch, HDFS-6982.v6.patch,
> HDFS-6982.v7.patch, nntop-design-v1.pdf
>
>
> In this jira we motivate the need for nntop, a tool that, similarly to what
> top does in Linux, gives the list of top users of the HDFS name node and
> gives insight about which users are sending majority of each traffic type to
> the name node. This information turns out to be the most critical when the
> name node is under pressure and the HDFS admin needs to know which user is
> hammering the name node and with what kind of requests. Here we present the
> design of nntop which has been in production at Twitter in the past 10
> months. nntop proved to have low cpu overhead (< 2% in a cluster of 4K
> nodes), low memory footprint (less than a few MB), and quite efficient for
> the write path (only two hash lookup for updating a metric).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
