Credit card monitor app: UI coming as empty after installing and starting demo services

[#1]

@Vadim Vaks

installlog.txtsimulatorlogs.txtstartdemoserviceslog.txtblankcctmsnapshot.jpg

I am getting empty page as shown in the attachment when I install and start the demo services of Credit card monitor app.

I also see that there is no process running on the port 8082 which is causing issues with simulator as well.

Attached are installer logs (installLog.txt), start demo services logs (startDemoServiceslog.txt) and simulator logs (simulatorLogs.txt).

Please go through the same and

[#2 · Accepted solution]

@Anil Reddy

You are actually very close. The UI is loading successfully. What you are seeing is an empty Inbox. Picture Outlook if there were no emails received. You will see Credit Card transactions coming through after you start the simulator.

You are correct that the reason the simulator cannot send messages is that there is nothing listening on 8082. Apache Nifi is supposed to have an instantiated flow that listens on 8082 for HTTP Post as JSON. That is how the simulator generates transactions for the application to process.

Looking at the install log, everything looks good except for when the process tried to instantiate the Nifi Flow. Do the following:

Log into Nifi (http://sandbox.hortonworks.com:9090/nifi), Capture a screen shot and respond to this answer with a comment that includes that screenshot.

[#3]

Dear @Vadim Vaks,

Thanks for the prompt reply. Please find the attached screenshot of Nifi in the attachment.nifi.jpg

[#4]

@Anil Reddy

Ok, the issue is that installer did not get the latest Nifi. I hate to ask... could you create a fresh sandbox and run the install.sh script grabbing the output as before. It should be grabbing the latest Nifi bits and it looks like it did not. API changed between Nifi 1.2 and Nifi 2.0 and that is why the flow did not instantiate. Try a fesh install, capture the logs like before and respond to this comment.

[#5]

@Vadim Vaks

I do tried that in the earlier stages. When I run installer.sh on sandbox without installing Nifi service, script is removing the existing version and getting the latest version of Nifi. But the problem we faced is that when Latest Nifi is installed, all other components going in to unrecoverable stale state (in ambari we see ? mark on all components and we can't resolve them).

Hence I installed Nifi service and start it which will skip getting the latest Nifi in the script. Probably I need to try installing Nifi 2.0 manually.

I also noticed that stop and start of HDP services from script are causing issues and making the components stale. I need to go to ambari and start them manually.

I would like to know if you are able to run installer.sh successfully on fresh sandbox of HDP 2.4 with out any alterations?

[#6]

@Anil Reddy

Looks like HDF 2.0 breaks HDP Ambari. In hindsight this makes sense since there was a lot of work done in HDF 2.0 to make it work well with it's own Ambari. The wrapper changed significantly. I put in some backward compatibility since Sandbox 2.4 already has a copy of HDF 1.2 available to install from Ambari. Give it another shot, it should work end to end now. One last thing, Storm 0.10.0 takes a while to fully load the topology. I have seen it take up to 5 minutes. Storm 1.0.1 is much more efficient and loads the same topology in a few seconds. Make sure that the Storm log shows that all bolts and spouts have fully loaded before you attempt to load the UI. This is because Storm is what initializes Hbase with the data required for the UI to start. You may see a stack trace related to Atlas, just ignore that. Let me know if that work

[#7]

@Anil Reddy

Glad to hear that the install worked. Would you mind accepting this answer as correct?

Without investigation, I would guess that Hbase and Metrics are going down due to memory pressure. Running the whole application inside the a VM puts a lot of pressure on your machine. Try stoping Metrics, Falcon, Flume services to save on memory.

[#8]

Never mind to accept the answer. I should appreciate your efforts.

You are correct, services are going down due to heavy load on the CPU. we are seeing CPU spike when the app is running and seeing the CPU usage in the range of 90 - 100%. We are working to bring the CPU usage down and your suggestion to stop using metrics, falcon and flume services helps.

Thanks.

[#9]

@Vadim Vaks

Thanks alot for the fixes. The script run fine fully on HDP 2.4 with out any changes and I am able to launch the app and see the UI.

FYI, when I left the app running for a while I am seeing that some of the services are going down (services that I have seen going down frequently are region server, hbase and metrics monitor).