Home | About | Sematext search-lucene.com search-hadoop.com
NEW: Monitor These Apps!
elasticsearch, apache solr, apache hbase, hadoop, redis, casssandra, amazon cloudwatch, mysql, memcached, apache kafka, apache zookeeper, apache storm, ubuntu, centOS, red hat, debian, puppet labs, java, senseiDB
 Search Hadoop and all its subprojects:

Switch to Plain View
Hive >> mail # user >> Re: dfs storage full on all slave machines of 6 machine hive cluster


+
Manish Bhoge 2013-03-18, 10:48
+
Zhiwen Sun 2013-03-18, 15:10
+
Chunky Gupta 2013-03-18, 15:58
Copy link to this message
-
Re: dfs storage full on all slave machines of 6 machine hive cluster
Look into your hdfs-site.xml & mapred-site.xml conf files.

*dfs.data.dir* propety contain your actual HDFS data path, better avoid
deleting anything from these directories.

*mapred.local.dir* contains temporary map-reduce job data, you can clean
this one.

"/mnt/hadoop-fs/dfs/data/current/" looks like your hdfs data path, this
mean your hive tables have grown to ~95% of your disk size. try deleting
hive tables or add more disk ( dropping a EXTERNAL hive table doesn't clear
the data from HDFS)

Thanks,

On Mon, Mar 18, 2013 at 9:28 PM, Chunky Gupta <[EMAIL PROTECTED]>wrote:

> Hi Zhiwen,
>
> /mnt/hadoop-fs/mapred/local/taskTracker/
>
> Inside this folder there are folders with different user name, can I
> delete these ?.
>
> I do not understand what this {*nouserdir*} you were talking about, can
> you please explain ?.
>
> Thanks,
> Chunky.
>
>
>
> On Mon, Mar 18, 2013 at 8:40 PM, Zhiwen Sun <[EMAIL PROTECTED]> wrote:
>
>> The folder "/mnt/hadoop-fs/dfs/data/current/" is the main folder of
>> datanode in hadoop.
>>
>> You can use *hadoop dfs -rmr {nouserdir} *to get more free space in HDFS.
>>
>> *Don't delete file directly in OS file system.*
>>
>> Zhiwen Sun
>>
>>
>>
>> On Mon, Mar 18, 2013 at 6:48 PM, Manish Bhoge <[EMAIL PROTECTED]
>> > wrote:
>>
>>> I think these directories belong to task tracker temporary storage. I am
>>> not very confident to conclude that go ahead with your clean up. So, wait
>>> for similar or an expert's response
>>>
>>> Sent from HTC via Rocket! excuse typo.
>>>
>>>  ------------------------------
>>> * From: * Chunky Gupta <[EMAIL PROTECTED]>;
>>> * To: * <[EMAIL PROTECTED]>;
>>> * Subject: * dfs storage full on all slave machines of 6 machine hive
>>> cluster
>>> * Sent: * Mon, Mar 18, 2013 10:37:39 AM
>>>
>>>   Hi,
>>>
>>> We have a 6 machine hive cluster. We are getting errors while a query is
>>> running and it fails. I found that on all 5 slaves storage is nearly full (
>>> 96%, 98%, 100%, 97%, 98% storage used) .
>>>
>>> On my slaves machines, this folder "/mnt/hadoop-fs/dfs/data/current/" is
>>> contributing 95% storage used. It contains folders with names "subdir0",
>>> "subdir1", etc and under them there are many files with name like
>>> "blk_-4071357924681234567" and blk_-4071357924681234567_246813.meta:, etc.
>>>
>>> I want to delete these subdir folders but I am not sure if it will not
>>> affects the tables which I have created.
>>>
>>> Can anyone help me and tell me what are these folders used for ?.
>>>
>>> Thanks,
>>> Chunky.
>>>
>>
>>
>
--
Alok Kumar
+
Chunky Gupta 2013-03-19, 10:36
+
Alok Kumar 2013-03-19, 18:32
+
Chunky Gupta 2013-03-18, 10:37
NEW: Monitor These Apps!
elasticsearch, apache solr, apache hbase, hadoop, redis, casssandra, amazon cloudwatch, mysql, memcached, apache kafka, apache zookeeper, apache storm, ubuntu, centOS, red hat, debian, puppet labs, java, senseiDB