Mailing-List: contact common-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: common-issues@hadoop.apache.org
Date: Tue, 22 Mar 2016 10:11:25 +0000 (UTC)
From: "Steve Loughran (JIRA)" <jira@apache.org>
To: common-issues@hadoop.apache.org
Message-ID: <JIRA.12952104.1458580978000.15591.1458641485773@Atlassian.JIRA>
In-Reply-To: <JIRA.12952104.1458580978000@Atlassian.JIRA>
References: <JIRA.12952104.1458580978000@Atlassian.JIRA>
 <JIRA.12952104.1458580978734@arcas>
Subject: [jira] [Commented] (HADOOP-12949) Add HTrace to the s3a connector
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable


    [ https://issues.apache.org/jira/browse/HADOOP-12949?page=3Dcom.atlassi=
an.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=3D15=
206107#comment-15206107 ]=20

Steve Loughran commented on HADOOP-12949:
-----------------------------------------

There's actually some metrics collection in openstack swift; look under {{o=
rg.apache.hadoop.fs.swift.util.DurationStats}} ; they log primarily to stdo=
ut, list min, max, (moving) arithmetic mean, stddev,, by HTTP verb.

# It's pretty low cost to do this; even when hbase sampling is inactive, th=
e stats for an FS can be collected.
# The stats showed that rackspace UK throttles delete requests; the more fi=
les in a directory I was cleaning up on teardown, the longer it took =E2=80=
=94only now exponentially, rather than linearly.
# I didn't hook the code up to the normal hadoop metrics; it's something I'=
d as an option now, because it does become something you need to monitor no=
w we are shifting to longer-lived applications.
# I'd add more on causes of operations, specifically: open(), seek(), durat=
ion of close(), delete() =E2=80=94things where the fact that object stores =
are generally O(files*data) means they don't work as expected ... finding t=
hat mismatch of expectations matters

More and more object stores are coming in. While s3 is the main one, it'd b=
e good to have the core stuff store neutral. The classes from hadoop-openst=
ack can be moved if that helps; the per-verb stuff is useful at the deep le=
vels, while htrace monitoring can track cost of specific actions.


> Add HTrace to the s3a connector
> -------------------------------
>
>                 Key: HADOOP-12949
>                 URL: https://issues.apache.org/jira/browse/HADOOP-12949
>             Project: Hadoop Common
>          Issue Type: Improvement
>            Reporter: Madhawa Gunasekara
>
> Hi All,=20
> s3, GCS, WASB, and other cloud blob stores are becoming increasingly impo=
rtant in Hadoop. But we don't have distributed tracing for these yet. It wo=
uld be interesting to add distributed tracing here. It would enable collect=
ing really interesting data like probability distributions of PUT and GET r=
equests to s3 and their impact on MR jobs, etc.
> I would like to implement this feature, Please shed some light on this=20
> Thanks,
> Madhawa


--
This message was sent by Atlassian JIRA
(v6.3.4#6332)