Bug 4178 – no job output when streaming with globusrun-ws

Bug 4178 - no job output when streaming with globusrun-ws

Summary:

no job output when streaming with globusrun-ws

Status:	RESOLVED FIXED

Product:	GRAM
Component:	wsrf gram clients
Version:	4.0.1
Platform:	IA64 Linux

Importance:	P3 major
Target Milestone:	4.0.2
Assigned To:	Peter Lane

URL:
Keywords:

Depends on:
Blocks:
	Show dependency tree / graph

Reported:	2006-01-26 17:04 by Stuart Martin
Modified:	2006-10-18 16:46 (History)

Attachments
Patch setting stdout/err RPs before going Active (1.37 KB, patch) 2006-02-01 13:41, Peter Lane	Details
Add an attachment (proposed patch, testcase, etc.)

Note

You need to log in before you can comment on or make changes to this bug.

Description From Stuart Martin 2006-01-26 17:04:14

On Teragrid, I submit a streaming job and there is no output displayed on the
terminal.  If I submit a job 
that echo info; sleep 30; echo info I can see the data in the stdout file in my
home dir, but the stdout data 
is never displayed to the terminal.  Looking at the gridftp log file, the only
commands that show up for the 
job are delete commands for the job's stdout and stderr.  So either the
globusrun-ws commands are 
failing (silently) or not being attempted for some reason.

Host: tg-grid1.uc.teragrid.org
GLOBUS_LOCATION: /soft/globus-4.0.1-r2/
container log file: /usr/local/logs/globus_container.tg-grid1.8443.log
Sample command: tg-grid1 /home/smartin> globusrun-ws -submit -s -c
/bin/hostname
globusrun-ws -dbg output: /home/smartin/stream_dbg_output

------- Comment #1 From Stuart Martin 2006-01-26 17:50:19 -------

Mike Showerman reports that it works fine on other sites.  Grid1 is sles8/ia64

------- Comment #2 From Stuart Martin 2006-01-27 11:18:09 -------

JP and I have reproduced this twice sofar.  After restarting the GT4 container,
the first job attempt 
works fine and the stdout data is displayed.  Subsequent jobs do not show the
stdout data.

smartin@tg-grid1:~> globusrun-ws -submit -s -c /bin/hostname
Delegating user credentials...Done.
Submitting job...Done.
Job ID: uuid:1a26ed0e-8f58-11da-98cd-0007e9d81215
Termination time: 01/28/2006 17:12 GMT
Current job state: Active
Current job state: CleanUp-Hold
tg-grid1
Current job state: CleanUp
Current job state: Done
Destroying job...Done.
Cleaning up any delegated credentials...Done.


smartin@tg-grid1:~> globusrun-ws -submit -s -c /bin/hostname
Delegating user credentials...Done.
Submitting job...Done.
Job ID: uuid:3da438fe-8f58-11da-b857-0007e9d81215
Termination time: 01/28/2006 17:13 GMT
Current job state: Active
Current job state: CleanUp-Hold
Current job state: CleanUp
Current job state: Done
Destroying job...Done.
Cleaning up any delegated credentials...Done.

------- Comment #3 From Joe Bester 2006-02-01 11:22:15 -------

I added some debug statements to globusrun-ws and it looks like sometimes the
stdout/stderr resource properties are not populated when globusrun-ws does
GetMultipleResourceProperties. Since neither stdout or stderr are returned,
streaming doesn't happen. Any idea why this is happening on the java side?

joe

------- Comment #4 From Peter Lane 2006-02-01 11:53:27 -------

If globusrun-ws tries to grab those RPs immediately after the job is submitted
then I could see this happening. I would suggest waiting until after a
notification occurs before trying to grab those values if this is the way
globusrun-ws behaves.

------- Comment #5 From Joe Bester 2006-02-01 11:56:14 -------

It checks the property after the ACTIVE response is returned from the
container.

joe

------- Comment #6 From Peter Lane 2006-02-01 13:41:07 -------

Created an attachment (id=835) [details]
Patch setting stdout/err RPs before going Active

Although I can't reproduce this myself, there is clearly a race condition here
since the Active state is set before the OpenStdout and OpenStderr internal
states are processed. I'm attaching a patch that simply moves the setting of
the state to the end of the processOpenStderrState() method in the state
machine.

------- Comment #7 From Peter Lane 2006-02-01 13:47:55 -------

The fix has been comitted to the globus_4_0_branch. If somebody who can
reliably
reproduce this problem could apply the patch and report back that would be
great.

------- Comment #8 From Peter Lane 2006-10-18 16:46:57 -------

Since there's been no response on this in terms of verifying the patch, I'm
resolving this as FIXED. Feel free to reopen if there's still a problem with
version 4.0.2 or above.