File a JIRA in issues.apache.org/jira under STORM describing what you want to
see with storm and the UI especially for trident. We are in the process of
merging with the JStorm project and they have completely redone how metrics
work, so with their new code it might be more doable.
- Bobby
On Monday, November 23, 2015 10:52 AM, Sachin Pasalkar
<[email protected]> wrote:
I also found that we can configure our own matrices to check how many tuples
are processed
http://www.bigdata-cookbook.com/post/72320512609/storm-metrics-how-to
Can’t we have this as part of trident doing internally?
From: Sachin Pasalkar <[email protected]>
Date: Monday, 23 November 2015 10:14 pm
To: "[email protected]" <[email protected]>, Bobby Evans
<[email protected]>
Subject: Re: Trident topology showing improper values in UI
Thanks Bobby for reply. However, I am aware of these things but my question
different, when topology is not processing message still count on UI increases
(Both on topology page/component page). Also, I have hardly seen the expected
count of message processed on component page.
From: Bobby Evans <[email protected]>
Reply-To: "[email protected]" <[email protected]>, Bobby Evans
<[email protected]>
Date: Monday, 23 November 2015 8:17 pm
To: "[email protected]" <[email protected]>
Subject: Re: Trident topology showing improper values in UI
The metrics are not always the most clear on the UI, and trident does not make
it any clearer. First of all there are several different modes on the UI. You
can look at the metrics over different time periods, 10 mins, 3 hours, 1 day,
and all time. If you have clicked on one of these links the page will display
numbers for just this range, except in the section labeled for specific time
periods. There is also a button that decides if we should include system stats
or not. If this is selected the aggregate numbers will include acks. If it is
not selected they will not.
The other thing to be aware of is that trident is a layer that sits on top of
storm so the UI displays what the trident topology was compiled down to, not as
much of what your code looks like. Trident spouts actually run in a bolt, but
control logic is in the actual spout. You can still pull out the number of
messages that a trident spout emitted in most cases, but you need to know where
to look, and you might need to click on the component page for the bolt running
the spout to be able to divide the counts up by which stream they were written
to. - Bobby
On Monday, November 23, 2015 7:38 AM, Sachin Pasalkar
<[email protected]> wrote:
Hi,
We are developing the trident topology, we observed that the acking in UI
increase continuously even though we do not process any message. I found
description what those acks
arehttps://github.com/miguno/kafka-storm-starter/issues/5. However, is there
any plan to show users processed message? As a user, I am not interested in
seeing these acks. I am very much interesting in acks for my messages.
Thanks,Sachin
