[
https://issues.apache.org/jira/browse/HDFS-7122?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14155248#comment-14155248
]
Tsz Wo Nicholas Sze commented on HDFS-7122:
-------------------------------------------
{quote}
The original patch was also reviewed for several weeks and given a +1 by Aaron,
who has been working full time on HDFS for several years. I also discussed the
design with Andrew offline (though I apologize for not commenting on the JIRA).
So, we had around 10 man-years of experience looking at the patch or its design
and we still missed the bug. I wouldn't say that "domain expertise" was the
issue there. Rather, it was a simple mistake which was quickly addressed after
being discovered.
{quote}
I am surprise that you are measuring the HDFS man-years and adding them up! Do
you think if it makes sense for 3 four-grade students to claim that they all
together is equivalent to a twelve-grade student?
I agree that it is a simple mistake (i.e. not a complicated one). However, I
do not understand why we did not revert the original patch but committed
another patch for reverting the code.
Please do post on JIRA if you have comments. Otherwise, it is hard for hard
people to work with you. Thanks.
{quote}
Nicholas – speaking of respect, could you please refrain from ad hominem
attacks like this in the JIRA?
{quote}
I was responding to Andrew's earlier comments about he working on CDH with
people in Cloudera but not working on Apache trunk with Apache open source
community. I am not sure why you have such interpretation.
{quote}
The data seems to show that Andrew works quite a bit with open source. In fact,
he is the number one most prolific committer to Hadoop in the last 6 months,
and the number two most prolific in the past year (#1 if you only count HDFS).
I don't think dragging unsubstantiated attacks on the open source commitment
level of different vendors is productive in the context of a bug fix.
{quote}
The man-years of experience counted earlier includes Andrew, Aaron and Todd.
The so called "most prolific in last 6 months" measurement includes only
Andrew. Is it suggesting that, for Andrew alone, the number of man-years is
very small and, for Andrew, Aaron and Todd, the "most prolific in last 6
months" measurement is not quite good?
> Use of ThreadLocal<Random> results in poor block placement
> ----------------------------------------------------------
>
> Key: HDFS-7122
> URL: https://issues.apache.org/jira/browse/HDFS-7122
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: namenode
> Affects Versions: 2.3.0
> Environment: medium-large environments with 100's to 1000's of DNs
> will be most affected, but potentially all environments.
> Reporter: Jeff Buell
> Assignee: Andrew Wang
> Priority: Blocker
> Labels: performance
> Fix For: 2.6.0
>
> Attachments: copies_per_slave.jpg,
> hdfs-7122-cdh5.1.2-testing.001.patch, hdfs-7122.001.patch
>
>
> Summary:
> Since HDFS-6268, the distribution of replica block copies across the
> DataNodes (replicas 2,3,... as distinguished from the first "primary"
> replica) is extremely poor, to the point that TeraGen slows down by as much
> as 3X for certain configurations. This is almost certainly due to the
> introduction of Thread Local Random in HDFS-6268. The mechanism appears to
> be that this change causes all the random numbers in the threads to be
> correlated, thus preventing a truly random choice of DN for each replica copy.
> Testing details:
> 1 TB TeraGen on 638 slave nodes (virtual machines on 32 physical hosts),
> 256MB block size. This results in 6 "primary" blocks on each DN. With
> replication=3, there will be on average 12 more copies on each DN that are
> copies of blocks from other DNs. Because of the random selection of DNs,
> exactly 12 copies are not expected, but I found that about 160 DNs (1/4 of
> all DNs!) received absolutely no copies, while one DN received over 100
> copies, and the elapsed time increased by about 3X from a pre-HDFS-6268
> distro. There was no pattern to which DNs didn't receive copies, nor was the
> set of such DNs repeatable run-to-run. In addition to the performance
> problem, there could be capacity problems due to one or a few DNs running out
> of space. Testing was done on CDH 5.0.0 (before) and CDH 5.1.2 (after), but I
> don't see a significant difference from the Apache Hadoop source in this
> regard. The workaround to recover the previous behavior is to set
> dfs.namenode.handler.count=1 but of course this has scaling implications for
> large clusters.
> I recommend that the ThreadLocal Random part of HDFS-6268 be reverted until a
> better algorithm can be implemented and tested. Testing should include a
> case with many DNs and a small number of blocks on each.
> It should also be noted that even pre-HDFS-6268, the random choice of DN
> algorithm produces a rather non-uniform distribution of copies. This is not
> due to any bug, but purely a case of random distributions being much less
> uniform than one might intuitively expect. In the above case, pre-HDFS-6268
> yields something like a range of 3 to 25 block copies on each DN.
> Surprisingly, the performance penalty of this non-uniformity is not as big as
> might be expected (maybe only 10-20%), but HDFS should do better, and in any
> case the capacity issue remains. Round-robin choice of DN? Better awareness
> of which DNs currently store fewer blocks? It's not sufficient that the total
> number of blocks is similar on each DN at the end, but that at each point in
> time no individual DN receives a disproportionate number of blocks at once
> (which could be a danger of a RR algorithm).
> Probably should limit this jira to tracking the ThreadLocal issue, and track
> the random choice issue in another one.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
