Yup. Away from /tmp
<property>
<name>dfs.name.dir</name>
<value>/hdfs_storage/name</value>
</property>
<property>
<name>dfs.data.dir</name>
<value>/hdfs_storage/data</value>
</property>
On Fri, Aug 5, 2016 at 4:42 PM, John Vines <[email protected]> wrote:
> How did you configure hdfs? Did you map the on disk directories away from
> tmp? The apache versions of hadoop point to /tmp (or at least used to) so
> when you reboot they get clobbered.
>
>
> On Fri, Aug 5, 2016 at 12:16 PM Michael Wall <[email protected]> wrote:
>
>> Ok, great. One node. In a production cluster, services are distributed
>> onto many nodes, so a random reboot has less impact. In your case, a
>> reboot means everything will die at the same time. Let's try this.
>>
>> - Get to a good state and insert some data into accumulo. Create table,
>> and run the insert command. Then call flush from the shell to persist the
>> data to disk.
>> - Scan the metadata table to see where the files are located in hdfs.
>> Maybe even hdfs df ls to see the files
>> - Reboot the VM
>> - Start up hdfs and zookeeper. Look for you files in hdfs. Assuming
>> they are there, start up Accumulo
>> - Where do you see "Waiting for accumulo to be initialized"
>> - Go to the accumulo monitor webpage, something like
>> http://localhost:50095. Click on the recent logs and let's see what
>> accumulo is saying.
>> - Log files will be $ACCUMULO_HOME/logs unless you reconfigured. There
>> should be log for the master process. What does that show?
>>
>> On Fri, Aug 5, 2016 at 12:06 PM, Kevin Cho <[email protected]> wrote:
>>
>>> Yes, I meant VM Reboot.
>>>
>>> On Fri, Aug 5, 2016 at 4:05 PM, Kevin Cho <[email protected]> wrote:
>>>
>>>> Hi Mike,
>>>>
>>>> Thanks for helping.
>>>> - What versions are you using?
>>>> Accumulo 1.7.2
>>>> Hadoop 2.7.2
>>>> ZooKeeper 3.4.8
>>>>
>>>> - By reboot, do you mean rebooted the box? How many nodes in your
>>>> cluster?
>>>> Just a single box. Single node.
>>>>
>>>> - How did you determine the init was successful?
>>>> I just run the 'accumulo init' after removing the /accumulo directory
>>>> in hdfs. Then I can start/stop the accumulo w/o any issue. It's only when
>>>> I start again after hard VM reboot.
>>>>
>>>> - What is currently happening on the master?
>>>> I'm not sure how to answer this. I'm very new to accumulo.
>>>>
>>>> - What messages are showing up on the monitor?
>>>> I'm not sure how to answer this. I'm very new to accumulo.
>>>>
>>>> - Can you send logs and jstacks?
>>>> I'm not sure how to answer this. I'm very new to accumulo.
>>>>
>>>> I just freshed installed those versions above on Ubuntu 14.04. I've
>>>> reinstalled it couple of times and still had the same result. It seems
>>>> unless I execute stop-all.sh before reboot, I'm not able to start accumulo
>>>> again after the reboot.
>>>>
>>>>
>>>>
>>>> On Fri, Aug 5, 2016 at 3:44 PM, Michael Wall <[email protected]> wrote:
>>>>
>>>>> Kevin,
>>>>>
>>>>> We are going to need more info. Here are some things I can thing of.
>>>>>
>>>>> - What versions are you using?
>>>>> - By reboot, do you mean rebooted the box? How many nodes in your
>>>>> cluster?
>>>>> - How did you determine the init was successful?
>>>>> - What is currently happening on the master?
>>>>> - What messages are showing up on the monitor?
>>>>> - Can you send logs and jstacks?
>>>>>
>>>>> Thanks,
>>>>>
>>>>> Mike
>>>>>
>>>>> On Fri, Aug 5, 2016 at 11:20 AM, Kevin Cho <[email protected]> wrote:
>>>>>
>>>>>> I was able to init accumulo and able to start/stop. However, if I
>>>>>> run the command 'reboot' and tried to start again.. I keep getting this
>>>>>>
>>>>>> Waiting for accumulo to be initialized
>>>>>>
>>>>>> It just keeps going and I'm not sure how to fix that. Of course, I
>>>>>> can re-init but I don't want to do that. The reason why I'm testing with
>>>>>> 'reboot' is that sometimes power gets rebooted or anyone w/ access can
>>>>>> reboot as well. Thanks for the help!
>>>>>>
>>>>>
>>>>>
>>>>
>>>
>>
