On Fri, Jul 9, 2010 at 9:37 AM, Jinsong Hu wrote:
> Hi, Todd:
> I tried it from scratch, It is still not good. Currently I am using yahoo
> hadoop 0.20.9 and hbase 0.20.5. I just have too much trouble with them. I am
> thinking of using a supported and blessed version of hadoop and hbase
> combin
riday, May 21, 2010 11:14 PM
To:
Subject: Re: Fw: hadoop data loss issue discovered. please fix!
Hi Jinsong,
I don't see any data loss here.
The sequence of events from the logs:
==> NN allocates block:
2010-05-18 21:21:29,731 INFO org.apache.hadoop.hdfs.StateChange: BLOCK*
NameSy
all our machine's time is sync'ed via ntp, so you can compare exactly side
> by side and show the
> sequence of events.
> The build I took is from yahoo 0.20.9 distribution.
>
> Jinsong
>
> --------------
> From: "Todd Lip
Hi Jinsong,
Could you upload a tarball of the log files somewhere from each of the DNs
and the RS involved? It's hard to trace through the log in the email (the
email added all kinds of wrapping, etc)
-Todd
On Fri, May 21, 2010 at 2:17 PM, Jinsong Hu wrote:
> Hi, There:
> While I used hadoop
Hi, There:
While I used hadoop 0.20.9-yahoo distribution and hbase 0.20.4 version, I
found that the hadoop lose blocks under certain
situation, and thus corrupt hbase tables.
I compared namenode, datanode and hbase regionserver and figured out the
reason.
The regionserver 10.110.8.85 asks n
