I really like this; I've been wondering about the easiest way to visualize the storm topology graph; especially since I'm a heavy trident user so my topologies are generated...with really long names for the bolts. What you're showing here solves that and adds operational metrics to boot, which is great. Have you considered building the backend on top of the existing storm UI app? You might be able to leverage existing code to talk to nimbus. It would be awesome if storm provided something like what you've demoed out of the box on the same server infrastructure.
On Thu, Mar 20, 2014 at 8:28 AM, Middleton, Jonathan < [email protected]> wrote: > Very pretty. Hope you carry this forward. > > > > You might consider sampling from Nimbus using Thrift. > > > > *From:* Klausen Schaefersinho [mailto:[email protected]] > *Sent:* Thursday, March 20, 2014 8:21 AM > *To:* [email protected] > *Subject:* Re: User Interface > > > > Hi, > > > > it is just a prototype yet with no backend... So i still have to decide > what backend to use... > > > > Cheers, > > > > Klaus > > > > On Thu, Mar 20, 2014 at 1:15 PM, Jean-Sebastien Vachon < > [email protected]> wrote: > > Looks very promising. I like the fact that you see the flow of tuples > between elements. > > > > Any idea when you will release this so that we can try it out with our > topology? > > > > *From:* Klausen Schaefersinho [mailto:[email protected]] > *Sent:* March-18-14 4:41 PM > *To:* [email protected] > *Subject:* User Interface > > > > Hi, > > > > I have been prototyping an alternative UI for monitoring a storm cluster. > The main driver was, that I would like to have a more condensed view of > what is going on in my cluster and that I would also like to monitor some > use case specific metrics, e.g. the training accuracy of a classifier or > so. I have put a dummy online: > > > > http://vommond.de/streamui/streamui.html > > > > http://vommond.de/streamui/streamui-dark.html > > > > The width and the color of the lines between the boxes correlate to the > number of events flowing between the spouts and bolts. The widgets in the > boxes show some other (dummy) metrics that one might of instance monitor. > You can customize the UI quite a lot, e.g. hide widgets or change their > type (but all changes are lost in the dummy). > > > > Before I move on, I have some questions: > > > > 1) For now I have not integrated the UI with any backend. So before I > start, I would be happy if > > anybody could give me some advice what kind of monitoring backend to use. > Ideally the backend would be: > > > > - Easy to integrate with storm and > > - Could also read the normal storm metrics. > > - Have custom metrics variables to monitor. > > > > Also I do not have too much time, so any kind of solution that has a REST > service and would not require any server side hacking would be a big plus. > > > > 2) I saw that quite some people are using Ganglia or Nagius, but I have > nether used it. Can any body share their experience? > > > > > > Off course all comments with regards to the UI are welcome. > > > > I am planing to release the source code in some days (after some cleaning > up), so any body is welcome to pick it up. > > > > > > Cheers, > > > > Klaus > > > > > ------------------------------ > > Aucun virus trouvé dans ce message. > Analyse effectuée par AVG - www.avg.fr > Version: 2014.0.4336 / Base de données virale: 3722/7203 - Date: 16/03/2014 > > >
