|
Our current Hadoop version in M3 is 0.20.205.0 after following the yum install steps for CentOS around March 25th. Anyone have any idea when 0.22.0 will be out via yum install or another process? We're facing some bz2 issues that have been fixed in 0.22. |
|
We are looking into this. Will update soon. Thank you.
(01 May '12, 19:23)
surenh
Actually, this JIRA has already been back-ported by MapR and is included in our current production release. We publish all the patches, including source code. For example, for 1.2.x you can find the list here: http://www.mapr.com/doc/display/MapR/Hadoop+Compatibility+in+Version+1.2
(01 May '12, 22:52)
tshiran ♦♦
We are running 1.2.3. We know that an issue exists because we have timeouts when the files are bz2 and it is very fast when they are not. Here is our version. maprcli dashboard info -version true
version
(02 May '12, 06:54)
surenh
Can you turn off MapR's compression on the directories where the bz2 files will be written/read? See the documentation on "hadoop mfs -setcompression off" to turn it off.
(02 May '12, 07:11)
MC Srivas ♦♦
|
|
What kinds of issues? Can you point us to the JIRAs that you are interested in? This is the main one right now that is causing timeout issues because of the way our clients are sending us data. It becomes a client onboarding issue.
(30 Apr '12, 15:49)
surenh
|
|
Have you tested the built-in compression that comes with MapR? We definitely want to use the built-in compression when we can. But in some situations, our clients send us bz2 files that are gigs in size. If we had to uncompress each of these before putting them into HDFS, our client onboarding time would increase significantly because of the size and number of such files. So in this case, we cannot practically use the built-in compression.
(30 Apr '12, 16:02)
surenh
Makes total sense.
(30 Apr '12, 16:04)
TedDunning ♦♦
|
