It can be confusing to figure out why an analysis failed or doesn't complete, because there are any number of reasons why this might happen. It could be that one of your input files is corrupted, wasn't fully uploaded to the DE, or is in the wrong format. Maybe the problem is with the analysis name. Or maybe there's a technical problem with the binary tool used for the app or with the Discovery Environment itself.
This page gives some tips on how to troubleshoot the problem so you can figure out whether it's a problem with your file or with the Discovery Environment, and then help you either work on getting the problem fixed or resubmitting another analysis. See Analyses FAQs as well.
Troubleshooting a failed or stalled analysis
If you know that an analysis typically completes in 20 minutes but you have one that still shows Running status 24 hours after you submitted it, if you didn't get any output files or the output files were not what you expected, there could be a problem with your input files, the parameters you used, or there could be a problem with the app.
For more information on analyses status help, see Using the Analyses Window - Viewing the status of your analysis and getting help.
1. Read the tool's documentation
CyVerse hosts hundreds of apps, and our staff are not experts on most of them. The best way to find out how an app should work is by reading the original documentation that came with the tool. You can access the documentation for an app by clicking on the "i" button next to the app name (see Viewing App and Tool Information). This will take you either directly to the tool documentation or to a wiki page that links out to the tool documentation. You should always understand how an app works before you use it to analyze data.
2. Check the log files
One of the main tools available for troubleshooting a failed analysis is the set of log files that are returned with each completed or failed analysis. These log files contain important information about the analysis, such as the settings that were used, files you used, and, in the case of a failed analysis, information to help explain why the analysis failed.
Because different apps are based on different tools, there is no standard method used for error reporting, so the same type of error may land in different log files. For example, one app may return errors to the stdout files (usually the screen, although it can be redirected and is generally captured in a log file here), while another saves its errors to the stderr files (which usually writes to a file, but can also be redirected). This means you may well have to look in more than one log file when troubleshooting a failed analysis.
The log files that most commonly contain error information are (numerals in the filename correspond to the step number that was logged in your analysis):
condor-stderr and condor-input-stdout log files contain errors and details about Condor, the batch manager program that handles the execution of your analyses submission in the analyses queue.
condor-input-stderr and condor-input-stdout files contain details about outputs from the tool upon which the app is based.
If there was a problem with your input files or parameters, this will often be reported in the condor-stderr file.
See About Log Files for a more complete description of the contents of the files that are returned, as well as useful tips for app integrators.
3. Check the input files and parameter
4. Did CyVerse go into maintenance?
If you had a long-running analysis that failed, make sure that CyVerse did not enter maintenance while it was running. Check the maintenance calendar.
If your job was interrupted by maintenance, you can relaunch the analysis using the same files and settings.
5. Check is the app is functioning properly
If you suspect that the app is not working correctly, try to run it with the example data. Most apps have a quickstart that lists example data in the List of Applications. You can also get to the quickstart by viewing the app info by clicking on the "i" button.
Requesting additional help
Please go through all the troubleshooting steps yourself before requesting help. The problem is often something that you can diagnose yourself.
If you still need help, you can submit a request for help directly in the Analyses window. The status of the analysis determines the Help information that is displayed.
In the Analyses window, find the analysis with the possible issue. Click the status (i.e. 'Failed') of the failed analysis whose outputs you want to view.
Review the troubleshooting suggestions above.
If you still need assistance, click I still need help and complete the form.
Getting additional help
View the CyVerse maintenance calendar for upcoming and current maintenance periods. You also can see the Status page on the CyVerse website and the most recent CyVerse Node newsletter (which lists the upcoming maintenance periods). Other helpful sources are the CyVerse Facebook page, CyVerse Tweets, or the CyVerse Google+ page, and within the DE you may receive system messages as well.