ArcGIS Server Data Extraction - Job does not exist or is inaccessible

Question

I have followed the documentation and video to set up a data extraction geoprocessing service, which I am accessing with the ArcGIS Viewer for Flex (v. 2.3.1). When I run the Extract Data widget in the viewer I get an error: Job '(job folder name)' does not exist or is inaccessible. When I go to the arcgisjobs folder on the server the folder is there, along with the scratch folder and the output .zip file containing the data that was clipped. According to the messages.xml file in the scratch folder, the job finishes successfully.

I have checked the ArcGIS Server and Windows logs and haven't found anything obvious that would point to an issue. Folder permissions should be correct, since the task can create the job folder, clip the data, and create the output .zip folder. The task will run successfully when used through ArcGIS Desktop.

So has anyone experienced similar behavior with the data extraction tool or any other GP tasks? Any ideas of where I can look next?

Environment:
ArcGIS Server 10, .NET, SP2 - Distributed installation
ArcGIS Desktop 10, SP2
Windows Server 2008 x64 R2

Thanks, Tim

Update:
I did some testing today (in Firefox) and found a few more things:

1. The GP Service does occasionally work, it seems. Every few tries I do get a prompt to download the extracted data, only to get a security sandbox violation:
http://SERVER IP/Samples/DataExtract/Flex/widgets/DataExtract/DataExtractWidget.swf cannot load data from http://SERVER NAME/arcgisjobs/sumner.dataextract_gpserver/j77643b1b114e458f82e6f416262aaf58/scratch/output.zip.

2. When I am prompted to download the file, Firebug shows that it cannot download my crossdomain.xml file. It shows the address as http://SERVER NAME/crossdomain.xml

In both cases, it is trying to access a http://servername URL, and can't get there. Problem is, I can't figure out where it's getting this URL. I tried modifying the virtual directory URL's and republishing the services using the ArcGIS Server IP instead of machine name, with no change in behavior.

Answer 1

If its the script I'm thinking of, the jobs directory is hardcoded in the script. Crack open the script and make sure the jobs directory in the script lines up with what you have in your environment.

From what you've said, it sounds like the job is completing successfully, it just is returning the wrong URL to the FlexApp to download the .zip file.

Answer 2

I have a AGS10 SP2 distributed setup that seems similar to yours. I've been developing a PDF service based on ArcPy.Mapping Simple ExportToPDF Geoprocessing Service Example and am encountering the same "Job xx does not exist or is inaccessible" message you are.

In my case, it seems to be an issue of timing: the task completes but the process tries to access the results before they can be moved from the SOC to the SOM's jobs folder.

A temporary workaround is to intercept the specific error, extract the JobID, and make a call to retrieve the generated PDF:

function gpError(e) {
    if (e.message.search("does not exist or is inaccessible") > -1) {
        var jobId = e.message.replace(/Job '([a-z,0-9]*)' does not exist or is inaccessible\./, "$1");
        gp.getResultData(jobId, "Output_PDF");
    }
    else {
        showDialog(dojo.string.substitute('<h1>Error: ${error}</h1>', { error: e.message }));
    }
}

This is definitely not something I would put into production but it does allow me to workaround the error and continue development work while I wait for ESRI to solve the problem.