Hi, We are still seeing this one.

Hello, I have another variation of this problem, I have a number of `pytest` tests, one of them will restart one of the system services (init scripts) on our test machines, when this test happens pytest and the parent bamboo bash script exit correctly but bamboo will wait until the system service that was restarted is terminated.

The service in question used the common `nohup` pattern and works correctly on may other CI systems and OSs

I've been able to condense the bug down to a single shell script example:

#!/bin/bash
set -e
cat > sleep.sh << EOF
echo starting
nohup /bin/sleep 500 </dev/null>/dev/null 2>&1 &
echo detached
EOF

cat > check.py << EOF
#!/usr/bin/python
"""
Test bamboo job termination. 
"""
import subprocess

def setup_module():
    print subprocess.check_output("/bin/sh", "sleep.sh"])

def test_foo():
    print "hello"

EOF

python -m py.test check.py
ps -ef

This should exit with success after about 5 seconds but will wait the full 500 seconds until the sleep executed by nohup has ended.

The problem does seem related in some way to STDIO file descriptors because if the subprocess.check_output() call is changed to subprocess.check_call() then bamboo does not block.
 

For Windows this issue can be solved with Groovy script. Groovy can execute an external program as an process:

"/bin/application.exe".execute()

And then check that application is running:

println "tasklist /fi \"imagename eq application.exe\"".execute().text

This can be achieved e.g. via

/bin/bash -c 'sleep 60  > /dev/null 2>&1 &'

NOTE: This is a severe security issue. A detached process can spy on subsequently executing plans, potentially obtaining on confidential source code, passwords exposed in commands, etc.

We have a requirement for this as well on windows. Any word on a time frame for a solution?

I also have this requirement from a windows perspective and have found no solutions so far.

I have the same requirement on windows, but can not get this working.

Weird. Following task/script configuration is working for me, that is: Bamboo executes the script content in the background and finishes the Job meanwhile.

Example I've used:

Script body: (using Script Task)

nohup /home/pstefaniak/sleep-10-minutes.sh > /home/pstefaniak/sleep.out 2> /home/pstefaniak/sleep.err < /dev/null &

script:
/home/pstefaniak/sleep-10-minutes.sh

#!/bin/sh
sleep 10m

What shall we do now? Maybe you should open Support Request via http://support.atlassian.com ? That way an Atlassian support engineer could assist you...

Thanks for the sugguestion of using nohup. However this still did not produce the desired result of going on to the next stage.

From my testing it appears that Bamboo is kicking the command or script off with an initial "sh" process and all scripts or commands there after are childs. The commands we tell to go to the background probably go to the background however Bamboo is waiting for the initial sh process to return. My guess is we need a way to send this sh command to the background.

Please give a test to a solution you think will work and let me know what you think.

Ryan,

You might be able to background the process (I'm not 100% sure as I didn't test it by myself) if you redirect the output streams and use nohup. e.g

nohup /tmp/boo.sh > foo.out 2> foo.err < /dev/null &

Could you please try using that in your script task?

regards,
PS