What is the best way to send failed job output from gearman worker?

Question

#!/usr/bin/env python
import sys
import json
import gearman
from fabric import *
from fabric.api import *
from gearman import GearmanWorker


#
# Run the ssh task
#
def exe_job(worker, job):
    d = json.loads(job.data)
    env.host_string = d['host'] 
    cmd = d['command']
    retries = int(d['retries'])
    output = ""

    # Run the fabric command. Do not abort on exit
    # Run job for the given number of retries
    tmp_retries = retries;
    while retries > 0:
        with settings(warn_only=True):
            result = run(cmd)
            output = output + str(result)

        if result.failed:
            if retries == 1:
                job.send_fail()
                break
            else:
                next
        else:
            break

        retries = retries - 1

    return output


#
# Main function
#
def main():
    gm_worker = gearman.GearmanWorker(['localhost:4730'])
    gm_worker.register_task('exe_job',exe_job) 
    gm_worker.work()


if __name__ == '__main__':
    main()

In my code, I am trying to retry a gearman job (which runs a fabric command) for the number of retries specified by the user. For every try, I am capturing and appending the output. In the last retry, if the job fails, I want to return the output back to the client.

What is happening is that job.send_fail() command simply breaks out and does not go to the "return output" command at all to return the output of the failed command.

Is there a better way to fail a job and also return the data to the client on exit/failure without the worker dying?

dennisyuan · Answer 1 · 2012-02-10T07:55:55.973

In Gearman, the send_fail() function takes no parameters, just tell the job server this job on this worker is failed, so the server can do retry or something else.

If you do your procedure synchronously, the better way should be do the "retry or not" work on the client side.

If you do your procedure asynchronously, I think you can use "sendException" function.(in the installed gearman module for python in my computer it's send_job_exception(). ) This function can take a exception data arguments for your client information.

And at last, you can just simply do it like this easy way: (but your gearman client will receive A "GEARMAN_SUCCESS" RETURN CODE!)

#some codes
while retries > 0:
    with settings(warn_only=True):
        result = run(cmd)
        output = output + str(result)

    if result.failed:
        if retries == 1:
            #job.send_fail()
            output = "FAILED_JOB" + "some_return_str"
            break
        else:
            next
    else:
        break

    retries = retries - 1

return output

this link will help you, too.

What is the best way to send failed job output from gearman worker?

1 Answers1