Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Chukwa >> mail # user >> Unable to start hicc

Copy link to this message
Re: Unable to start hicc
Cluster name is found from table: SystemMetrics column family: system
column: ctag
The data is populated by Chukwa agent.  In the Chukwa agent configuration,
we put extra tag cluster="demo" for the default configuration.

It may be possible that SystemMetrics table doesn't have column family:


On Thu, May 24, 2012 at 5:21 AM, Prakhar Srivastava

> Hi Eric,
> The ClusterSummary table is already populated. I have once previously run
> the Pig Job and it has successfully populated the ClusterSummary table.
> The problem was with the HICC application not being able to load the
> cluster name. I did a small workaround to find if it was the problem. I
> modified the cluster variable in the index.jsp in the running HICC
> application to the cluster name provided in the chukwa agent configuration.
> This time, all the charts got populated. I am still not sure where does
> HICC pick up the cluster name on first run.
> Regards,
> Prakhar
> On Thu, May 24, 2012 at 11:47 AM, Eric Yang <[EMAIL PROTECTED]> wrote:
>> Yes, this means the ClusterSummary table has not been populated.  You
>> need to run Apache Pig with ClusterSummary.pig script to aggregate the
>> data.  Without doing that, you can try to use Graph Explorer to see what is
>> collected as raw data on HBase.
>> regards,
>> Eric
>> On Tue, May 22, 2012 at 11:49 PM, Prakhar Srivastava <
>> [EMAIL PROTECTED]> wrote:
>>> Hi Eric,
>>> While I check on your points,I found that the Cluster selector drop down
>>> show no cluster. The drop down menu is empty. Could this be an issue???
>>> Thanks,
>>> Prakhar
>>> On Wed, May 23, 2012 at 12:03 PM, Eric Yang <[EMAIL PROTECTED]> wrote:
>>>> Hi Prakhar,
>>>> If you have data in cluster summary, this means the system is working
>>>> as expected and data analytics script is aggregating fine.  Could it be
>>>> timezone issue?  Chukwa expects the server to run on UTC time zone to
>>>> prevent day light saving time to have over lapping data.  Javascript layer
>>>> translates UTC time to local timezone.  If the server is running in a
>>>> different timezone, it could offset the UI display time by n number of
>>>> hours.  You can try to look at data from last 7 days to see if any data is
>>>> getting populated?  You can also use graph explorer to browse the data in
>>>> HBase.  Hope this helps.
>>>> regards,
>>>> Eric
>>>> On Tue, May 22, 2012 at 11:22 PM, Prakhar Srivastava <
>>>> [EMAIL PROTECTED]> wrote:
>>>>> Hi Eric,
>>>>> I removed the disabled directory. I also had to modify the
>>>>> utilization.descriptor file because of some mismatched parenthesis.
>>>>> Now, the HICC dashboard loads fine. However, the charts are not
>>>>> getting populated. Only the Job summary widget lists the recently run jobs
>>>>> which concludes that HICC is able to talk to HBase. I can view the contents
>>>>> in the HBase tables like ClusterSummary using the HBase shell.
>>>>> Can you suggest how can I debug this since I am not getting any
>>>>> exceptions in the HICC logs.
>>>>> Thanks,
>>>>> Prakhar
>>>>> On Tue, May 22, 2012 at 9:01 PM, Eric Yang <[EMAIL PROTECTED]> wrote:
>>>>>> Hi Prakhar,
>>>>>> Try to remove the disabled directory from widget directory and
>>>>>> restart hicc.  It looks like the widgetbean is throwing null pointer
>>>>>> exception when attempt to serialized the disabled directory.  I remember to
>>>>>> put a fix for this in trunk, but it may not have been in 0.5.
>>>>>> Sent from my iPhone
>>>>>> On May 22, 2012, at 8:24 AM, Prakhar Srivastava <
>>>>>> [EMAIL PROTECTED]> wrote:
>>>>>> Hi Eric,
>>>>>> Thanks for the response.
>>>>>> The hicc running user is the hadoop user. So read/write permissions
>>>>>> are fine. I anyways changed hicc permissions to 777.
>>>>>> *The hicc directory looks like:*
>>>>>> *Name*
>>>>>> *Type*
>>>>>> *Size*
>>>>>> *Replication*