You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@accumulo.apache.org by Kesten Broughton <kb...@21ct.com> on 2014/01/26 22:45:16 UTC
/accumulo dir remains after reformatting name node and deleting
datanodes
>
>We did experiments using the root dir for both name nodes and datanodes.
>Now we want to wipe out the old hdfs cluster, keeping the name node at
>/hadoop/hdfs/namenode but moving the datanode fs to /dev/sdb etc.
After removing /hadoop/hdfs/data on each of the datanode machines and doing
hadoop namenode -format
I still see /accumulo on
hadoop fs -ls /
I was expecting it to be gone.
kesten
Re: /accumulo dir remains after reformatting name node and deleting datanodes
Posted by Sean Busbey <bu...@cloudera.com>.
HDFS will definitely not function without the NameNode. I'm not sure how
the NameNode process will respond if you delete its storage directory out
from under it (that is, while it is running).
What's the output of `alternatives --display hadoop-conf`?
Is txoig-hdfs01 the name of your namenode server?
How are you shutting down HDFS before deleting things?
On Sun, Jan 26, 2014 at 5:10 PM, Kesten Broughton <kb...@21ct.com>wrote:
> Apparently, there’s something i really don’t understand about hdfs data
> files.
> I assumed that without the name node, hdfs would cease to function, but
> hadoop fs -ls / still shows stuff when i remove the name node.
>
> [root@txoig-hdfs01:~]$ rm -Rf /hadoop/hdfs/namenode
> [root@txoig-hdfs01:~]$ su hdfs
> [hdfs@txoig-hdfs01:root]$ hadoop fs -ls /
> Found 5 items
> drwxr-xr-x - hdfs hdfs 0 2014-01-26 13:36 /accumulo
> drwxr-xr-x - hdfs hdfs 0 2013-12-19 18:17 /apps
> drwx------ - mapred hdfs 0 2014-01-26 16:39 /mapred
> drwxrwxrwx - hdfs hdfs 0 2014-01-26 14:56 /tmp
> drwxr-xr-x - hdfs hdfs 0 2013-12-30 15:11 /user
> [hdfs@txoig-hdfs01:root]$ ls /hadoop/hdfs/
> total 8.0K
> drwxr-xr-x. 2 root 4.0K Jan 26 17:02 .
> drwxr-xr-x. 5 root 4.0K Dec 19 18:13 ..
>
>
> [root@txoig-hdfs02:/]$ rm -Rf /hadoop/hdfs/namesecondary
> [root@txoig-hdfs02:/]$ ls /hadoop/hdfs/
> total 8.0K
> drwxr-xr-x. 2 root 4.0K Jan 26 17:04 .
> drwxr-xr-x. 6 root 4.0K Dec 19 18:14 ..
> [root@txoig-hdfs02:/]$ hadoop fs -ls /
> Found 5 items
> drwxr-xr-x - hdfs hdfs 0 2014-01-26 13:36 /accumulo
> drwxr-xr-x - hdfs hdfs 0 2013-12-19 18:17 /apps
> drwx------ - mapred hdfs 0 2014-01-26 16:39 /mapred
> drwxrwxrwx - hdfs hdfs 0 2014-01-26 14:56 /tmp
> drwxr-xr-x - hdfs hdfs 0 2013-12-30 15:11 /user
>
> [root@txoig-hdfs03:/]$ ls /hadoop/hdfs/
> total 8.0K
> drwxr-xr-x. 2 root 4.0K Jan 26 16:23 .
> drwxr-xr-x. 6 root 4.0K Dec 19 18:14 ..
>
> cat /etc/hadoop/conf.empty/hdfs-site.xml contains
>
>
> <property>
> <name>dfs.name.dir</name>
> <value>/hadoop/hdfs/namenode</value>
> </property>
>
>
> [root@txoig-hdfs03:/]$ ls /etc/hadoop
> total 12K
> drwxr-xr-x. 3 root 4.0K Dec 19 18:10 .
> drwxr-xr-x. 98 root 4.0K Jan 26 16:04 ..
> lrwxrwxrwx. 1 hdfs 29 Dec 19 18:10 conf -> /etc/alternatives/hadoop-conf
> drwxr-xr-x. 2 root 4.0K Jan 26 16:55 conf.empty
>
>
>
>
> On 1/26/14, 4:37 PM, "Kesten Broughton" <kb...@21ct.com> wrote:
>
> >It¹s possible something was running. I only did an accumulo/stop-all.sh
> >and sometimes i¹ve noticed i have to go to each node to stop-here.sh.
> >
> >[root@txoig-hdfs02:accumulo-1.5.0]$ ls /hadoop/hdfs
> >drwxr-xr-x. 5 hdfs 4.0K Jan 26 16:10 namesecondary
> >
> >
> >[root@txoig-hdfs01:~]$ ls /hadoop/hdfs/
> >drwxr-xr-x. 5 hdfs 4.0K Jan 26 16:10 namenode
> >
> >
> >[root@txoig-hdfs03:~]$ ls /hadoop/hdfs
> >drwxr-xr-x. 4 root 4.0K Jan 26 16:13 namenode
> ># this is weird. Wasn¹t expecting a second name node. We aren¹t doing HA.
> >
> >
> >But shutting down all accumulo and hdfs and then doing the same procedure,
> >i still see
> >
> >[root@txoig-hdfs01:~]$ hadoop fs -ls /
> >Found 5 items
> >drwxr-xr-x - hdfs hdfs 0 2014-01-26 13:36 /accumulo
> >drwxr-xr-x - hdfs hdfs 0 2013-12-19 18:17 /apps
> >drwx------ - mapred hdfs 0 2014-01-26 14:32 /mapred
> >drwxrwxrwx - hdfs hdfs 0 2014-01-26 14:56 /tmp
> >drwxr-xr-x - hdfs hdfs 0 2013-12-30 15:11 /user
> >
> >
> >
> >On 1/26/14, 4:16 PM, "Christopher" <ct...@apache.org> wrote:
> >
> >>Was any part of Hadoop or Accumulo still running when you did this?
> >>Are you sure you deleted the right HDFS data directory referenced in
> >>your Hadoop configuration?
> >>
> >>--
> >>Christopher L Tubbs II
> >>http://gravatar.com/ctubbsii
> >>
> >>
> >>On Sun, Jan 26, 2014 at 4:45 PM, Kesten Broughton <kb...@21ct.com>
> >>wrote:
> >>>>
> >>>>We did experiments using the root dir for both name nodes and
> >>>>datanodes.
> >>>>Now we want to wipe out the old hdfs cluster, keeping the name node at
> >>>>/hadoop/hdfs/namenode but moving the datanode fs to /dev/sdb etc.
> >>>
> >>> After removing /hadoop/hdfs/data on each of the datanode machines and
> >>>doing
> >>> hadoop namenode -format
> >>>
> >>> I still see /accumulo on
> >>> hadoop fs -ls /
> >>>
> >>> I was expecting it to be gone.
> >>>
> >>> kesten
> >>>
> >
>
>
Re: /accumulo dir remains after reformatting name node and deleting
datanodes
Posted by Kesten Broughton <kb...@21ct.com>.
Apparently, there’s something i really don’t understand about hdfs data
files.
I assumed that without the name node, hdfs would cease to function, but
hadoop fs -ls / still shows stuff when i remove the name node.
[root@txoig-hdfs01:~]$ rm -Rf /hadoop/hdfs/namenode
[root@txoig-hdfs01:~]$ su hdfs
[hdfs@txoig-hdfs01:root]$ hadoop fs -ls /
Found 5 items
drwxr-xr-x - hdfs hdfs 0 2014-01-26 13:36 /accumulo
drwxr-xr-x - hdfs hdfs 0 2013-12-19 18:17 /apps
drwx------ - mapred hdfs 0 2014-01-26 16:39 /mapred
drwxrwxrwx - hdfs hdfs 0 2014-01-26 14:56 /tmp
drwxr-xr-x - hdfs hdfs 0 2013-12-30 15:11 /user
[hdfs@txoig-hdfs01:root]$ ls /hadoop/hdfs/
total 8.0K
drwxr-xr-x. 2 root 4.0K Jan 26 17:02 .
drwxr-xr-x. 5 root 4.0K Dec 19 18:13 ..
[root@txoig-hdfs02:/]$ rm -Rf /hadoop/hdfs/namesecondary
[root@txoig-hdfs02:/]$ ls /hadoop/hdfs/
total 8.0K
drwxr-xr-x. 2 root 4.0K Jan 26 17:04 .
drwxr-xr-x. 6 root 4.0K Dec 19 18:14 ..
[root@txoig-hdfs02:/]$ hadoop fs -ls /
Found 5 items
drwxr-xr-x - hdfs hdfs 0 2014-01-26 13:36 /accumulo
drwxr-xr-x - hdfs hdfs 0 2013-12-19 18:17 /apps
drwx------ - mapred hdfs 0 2014-01-26 16:39 /mapred
drwxrwxrwx - hdfs hdfs 0 2014-01-26 14:56 /tmp
drwxr-xr-x - hdfs hdfs 0 2013-12-30 15:11 /user
[root@txoig-hdfs03:/]$ ls /hadoop/hdfs/
total 8.0K
drwxr-xr-x. 2 root 4.0K Jan 26 16:23 .
drwxr-xr-x. 6 root 4.0K Dec 19 18:14 ..
cat /etc/hadoop/conf.empty/hdfs-site.xml contains
<property>
<name>dfs.name.dir</name>
<value>/hadoop/hdfs/namenode</value>
</property>
[root@txoig-hdfs03:/]$ ls /etc/hadoop
total 12K
drwxr-xr-x. 3 root 4.0K Dec 19 18:10 .
drwxr-xr-x. 98 root 4.0K Jan 26 16:04 ..
lrwxrwxrwx. 1 hdfs 29 Dec 19 18:10 conf -> /etc/alternatives/hadoop-conf
drwxr-xr-x. 2 root 4.0K Jan 26 16:55 conf.empty
On 1/26/14, 4:37 PM, "Kesten Broughton" <kb...@21ct.com> wrote:
>It¹s possible something was running. I only did an accumulo/stop-all.sh
>and sometimes i¹ve noticed i have to go to each node to stop-here.sh.
>
>[root@txoig-hdfs02:accumulo-1.5.0]$ ls /hadoop/hdfs
>drwxr-xr-x. 5 hdfs 4.0K Jan 26 16:10 namesecondary
>
>
>[root@txoig-hdfs01:~]$ ls /hadoop/hdfs/
>drwxr-xr-x. 5 hdfs 4.0K Jan 26 16:10 namenode
>
>
>[root@txoig-hdfs03:~]$ ls /hadoop/hdfs
>drwxr-xr-x. 4 root 4.0K Jan 26 16:13 namenode
># this is weird. Wasn¹t expecting a second name node. We aren¹t doing HA.
>
>
>But shutting down all accumulo and hdfs and then doing the same procedure,
>i still see
>
>[root@txoig-hdfs01:~]$ hadoop fs -ls /
>Found 5 items
>drwxr-xr-x - hdfs hdfs 0 2014-01-26 13:36 /accumulo
>drwxr-xr-x - hdfs hdfs 0 2013-12-19 18:17 /apps
>drwx------ - mapred hdfs 0 2014-01-26 14:32 /mapred
>drwxrwxrwx - hdfs hdfs 0 2014-01-26 14:56 /tmp
>drwxr-xr-x - hdfs hdfs 0 2013-12-30 15:11 /user
>
>
>
>On 1/26/14, 4:16 PM, "Christopher" <ct...@apache.org> wrote:
>
>>Was any part of Hadoop or Accumulo still running when you did this?
>>Are you sure you deleted the right HDFS data directory referenced in
>>your Hadoop configuration?
>>
>>--
>>Christopher L Tubbs II
>>http://gravatar.com/ctubbsii
>>
>>
>>On Sun, Jan 26, 2014 at 4:45 PM, Kesten Broughton <kb...@21ct.com>
>>wrote:
>>>>
>>>>We did experiments using the root dir for both name nodes and
>>>>datanodes.
>>>>Now we want to wipe out the old hdfs cluster, keeping the name node at
>>>>/hadoop/hdfs/namenode but moving the datanode fs to /dev/sdb etc.
>>>
>>> After removing /hadoop/hdfs/data on each of the datanode machines and
>>>doing
>>> hadoop namenode -format
>>>
>>> I still see /accumulo on
>>> hadoop fs -ls /
>>>
>>> I was expecting it to be gone.
>>>
>>> kesten
>>>
>
Re: /accumulo dir remains after reformatting name node and deleting
datanodes
Posted by Kesten Broughton <kb...@21ct.com>.
It¹s possible something was running. I only did an accumulo/stop-all.sh
and sometimes i¹ve noticed i have to go to each node to stop-here.sh.
[root@txoig-hdfs02:accumulo-1.5.0]$ ls /hadoop/hdfs
drwxr-xr-x. 5 hdfs 4.0K Jan 26 16:10 namesecondary
[root@txoig-hdfs01:~]$ ls /hadoop/hdfs/
drwxr-xr-x. 5 hdfs 4.0K Jan 26 16:10 namenode
[root@txoig-hdfs03:~]$ ls /hadoop/hdfs
drwxr-xr-x. 4 root 4.0K Jan 26 16:13 namenode
# this is weird. Wasn¹t expecting a second name node. We aren¹t doing HA.
But shutting down all accumulo and hdfs and then doing the same procedure,
i still see
[root@txoig-hdfs01:~]$ hadoop fs -ls /
Found 5 items
drwxr-xr-x - hdfs hdfs 0 2014-01-26 13:36 /accumulo
drwxr-xr-x - hdfs hdfs 0 2013-12-19 18:17 /apps
drwx------ - mapred hdfs 0 2014-01-26 14:32 /mapred
drwxrwxrwx - hdfs hdfs 0 2014-01-26 14:56 /tmp
drwxr-xr-x - hdfs hdfs 0 2013-12-30 15:11 /user
On 1/26/14, 4:16 PM, "Christopher" <ct...@apache.org> wrote:
>Was any part of Hadoop or Accumulo still running when you did this?
>Are you sure you deleted the right HDFS data directory referenced in
>your Hadoop configuration?
>
>--
>Christopher L Tubbs II
>http://gravatar.com/ctubbsii
>
>
>On Sun, Jan 26, 2014 at 4:45 PM, Kesten Broughton <kb...@21ct.com>
>wrote:
>>>
>>>We did experiments using the root dir for both name nodes and datanodes.
>>>Now we want to wipe out the old hdfs cluster, keeping the name node at
>>>/hadoop/hdfs/namenode but moving the datanode fs to /dev/sdb etc.
>>
>> After removing /hadoop/hdfs/data on each of the datanode machines and
>>doing
>> hadoop namenode -format
>>
>> I still see /accumulo on
>> hadoop fs -ls /
>>
>> I was expecting it to be gone.
>>
>> kesten
>>
Re: /accumulo dir remains after reformatting name node and deleting datanodes
Posted by Christopher <ct...@apache.org>.
Was any part of Hadoop or Accumulo still running when you did this?
Are you sure you deleted the right HDFS data directory referenced in
your Hadoop configuration?
--
Christopher L Tubbs II
http://gravatar.com/ctubbsii
On Sun, Jan 26, 2014 at 4:45 PM, Kesten Broughton <kb...@21ct.com> wrote:
>>
>>We did experiments using the root dir for both name nodes and datanodes.
>>Now we want to wipe out the old hdfs cluster, keeping the name node at
>>/hadoop/hdfs/namenode but moving the datanode fs to /dev/sdb etc.
>
> After removing /hadoop/hdfs/data on each of the datanode machines and doing
> hadoop namenode -format
>
> I still see /accumulo on
> hadoop fs -ls /
>
> I was expecting it to be gone.
>
> kesten
>