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

Switch to Plain View
Accumulo, mail # dev - Monitor: Adding graphs for each table and server


+
Supun Kamburugamuva 2013-04-05, 15:21
+
Eric Newton 2013-04-05, 16:46
+
Supun Kamburugamuva 2013-04-05, 17:29
+
Eric Newton 2013-04-05, 17:34
+
Supun Kamburugamuva 2013-04-10, 15:11
+
Eric Newton 2013-04-10, 15:33
+
Josh Elser 2013-04-10, 15:51
+
Keith Turner 2013-04-05, 16:20
Copy link to this message
-
Re: Monitor: Adding graphs for each table and server
Billie Rinaldi 2013-04-05, 16:19
On Fri, Apr 5, 2013 at 8:21 AM, Supun Kamburugamuva <[EMAIL PROTECTED]>wrote:

> Hi all,
>
> According to [1] we need to add some graphs that displays statistics for
> each table and server in Accumulo. The current overview page in the monitor
> displays a summery (average) across all these entities.
>
> I can see the overview page is generated in TServersServlet where it pull
> statistics to display.
>
> For adding these new statistics should we create a new page that displays
> table statistics? If we display them in the current overview page, I
> suppose it will have lot of graphs.
>

Hi Supun!

I think the overview page is generated by the DefaultServlet.

There are already per-table pages generated by the TablesServlet, so
perhaps the table graphs can be added there.  If you click on "Tables" on
the left navigation bar and then click on a particular table name, it takes
you to a per-table view.  Right now the page just shows a table of stats
broken down by tablet server.

There are also per-tablet server pages generated by TServersServlet, if you
click on Tablet Servers on the left and then a particular server name.
Those have a bunch of information on them, but still no graphs.
>
> ACCUMULO-594 <https://issues.apache.org/jira/browse/ACCUMULO-594> talks
> about OpenTSBD. Can you please give me some references on why we need
> OpenTSBD?
>

Currently the monitor process keeps all of the data needed for the plots in
memory.  If we start doing a lot more plots, there would probably be too
much data to keep it all in memory.  The obvious place to store the data
would be in Accumulo, and OpenTSBD is one method of storing time-series
data in Accumulo.  There are other methods as well, but it would be worth
evaluating whether using OpenTSDB would save us some work.  Eric has done
some experimenting with OpenTSDB and Accumulo: http://s.apache.org/Jzx

Billie
>
> Thanks,
> Supun..
>
> [1] https://issues.apache.org/jira/browse/ACCUMULO-594
>