Alright.
Its getting weird now, but I somehow made it work. Probably I am having
some messed up configuration.
First of all, I start and stop the Jobtracker etc by executing this on
shell:
for service in /etc/init.d/hadoop-0.20-*; do sudo $service stop; done
They are actually already started when I boot my computer.
Second: I had a typo in my hadoop-env.sh in the part, where one can set
the "export HADOOP_CLASSPATH" and referencing to the hbase lib folder.
But I think that was not the problem.
Third: in my hbase-env.sh, i moved
"export HBASE_LIBRARY_PATH=/usr/lib/hadoop/lib/native/Linux-amd64-64"
to the beginning of the file. But I think that was not the problem
either, since it should not be necessary if one installs it using the
repository.
Forth:
I always used to start hbase by using this:
sudo /usr/lib/hbase/bin/start-hbase.sh
BUT now I tried on shell:
sudo hbase master start
and now it works an I can create tables with snappy compression. And I
am totally confused now about the commands to use. Doesn't
"start-hbase.sh" doing exactly the same thing?
Regards,
Christopher
Am 24.11.2011 16:52, schrieb Lars George:
Yeah, I meant the packages, and in which order they were installed, combined with when
the service was started. The "hbase" command loads the same settings as the
daemon, since it is started using the same scripts. The only explanation I have is that
the daemon has an outdated environment, due to it already running. Have you tried
restarting them?
Lars
On Nov 24, 2011, at 16:31, Christopher Dorner<[email protected]>
wrote:
Yes, snappy is installed.
Lars, do you mean with "the rest" the hadoop datanode, namenode,
jobtracker, hmaster, etc.?
I am not really sure, but I think I stopped the services first, before
installing hadoop-0.20-native package.
Christopher
Am 24.11.2011 12:47, schrieb Lars George:
He has Snappy, or else the CompressionTest would not work.
I can only imagine that there is an issue with the RS process not being able to
access the library path. Have you installed the hadoop-0.20-native package
while the rest was already running?
Lars
On Nov 24, 2011, at 12:39 PM, Gaojinchao wrote:
It seems you need install snappy.
-----邮件原件-----
发件人: Christopher Dorner [mailto:[email protected]]
发送时间: 2011年11月24日 18:10
收件人: [email protected]
主题: Error creating table using snappy compression
Hi,
i posted this question already on the cloudera list, but i didn't get a
solution yet. So i want to ask here again.
I am currently running Hadoop and HBase in pseudo-distributed mode
using CDH3-u2. In this update, snappy was included for HBase 0.90.4-
cdh3u2.
I wanted to try it out and compare size and speed to lzo (which works
fine). But when i try to create a table, or alter an existing table, i
get an error.
Here is what i do on HBase shell (same effect when i use java API).
create 'testsnappy', {NAME => 'f1', COMPRESSION => 'SNAPPY'}
ERROR: org.apache.hadoop.hbase.client.RegionOfflineException: Only 0
of 1 regions are online; retries exhausted.
On the HBase Master Website in can see the new table in the "Regions
in Transition" section. The state is altering between "opening" and
"closed".
The log is full of those:
ERROR org.apache.hadoop.hbase.regionserver.handler.OpenRegionHandler:
Failed open of region=testsnappy,,
1321955383777.7e5e71006335788551c2d6e90d5f9dee.
java.io.IOException: Compression algorithm 'snappy' previously failed
test.
at
org.apache.hadoop.hbase.util.CompressionTest.testCompression(CompressionTest.java:
78)
at
org.apache.hadoop.hbase.regionserver.HRegion.checkCompressionCodecs(HRegion.java:
2670)
at
org.apache.hadoop.hbase.regionserver.HRegion.openHRegion(HRegion.java:
2659)
at
org.apache.hadoop.hbase.regionserver.HRegion.openHRegion(HRegion.java:
2647)
at
org.apache.hadoop.hbase.regionserver.handler.OpenRegionHandler.openRegion(OpenRegionHandler.java:
312)
at
org.apache.hadoop.hbase.regionserver.handler.OpenRegionHandler.process(OpenRegionHandler.java:
99)
at
org.apache.hadoop.hbase.executor.EventHandler.run(EventHandler.java:
158)
at java.util.concurrent.ThreadPoolExecutor
$Worker.runTask(ThreadPoolExecutor.java:886)
at java.util.concurrent.ThreadPoolExecutor
$Worker.run(ThreadPoolExecutor.java:908)
at java.lang.Thread.run(Thread.java:662)
It indicates a failed snappy test, but when i run the test from a
shell:
hbase org.apache.hadoop.hbase.util.CompressionTest afolder snappy
11/11/22 10:57:17 WARN snappy.LoadSnappy: Snappy native library is
available
11/11/22 10:57:17 INFO util.NativeCodeLoader: Loaded the native-hadoop
library
11/11/22 10:57:17 INFO snappy.LoadSnappy: Snappy native library loaded
11/11/22 10:57:17 INFO compress.CodecPool: Got brand-new compressor
SUCCESS
So it seems to be successful.
What am i missing? Is this just related to pseudo-dstributed mode?
I am going to run MapReduce Jobs on HBase tables probably from next
week on a real cluster (also using CDH3-u2) and i would like to avoid
these problems then :)
When i set snappy as compression for Mapper outputs, i don't get any
errors and the jobs run fine.
conf.set("mapred.compress.map.output", "true");
conf.set("mapred.map.output.compression.codec",
"org.apache.hadoop.io.compress.SnappyCodec");
Thanks for help,
Christopher
