Skip to content

Custom Hooks

Life cycle hooks are the script used to start / stop / monitor your apps by Brainlife. By default, it looks for executable installed on each resource in the PATH with named start, status, and stop. Resource owner needs to make sure these scripts are installed and accessible by your apps.

For most PBS, SLURM, and vanilla VM, resource owner can install ABCD default hooks.

By default, the start hook should look for a file named main to start your app. Therefore, the only file required to make your app runnable by Brainlife is this main executable on the root directory of the app's git repository.

Under most circumstances, app developers shouldn't have to worry about these hook scripts. However, if your app requires some special mechanism to start / stop and monitor your app, you might need to provide your own hook scripts.

You can specify the paths to these hook scripts by creating a file named package.json

1
2
3
4
5
6
7
{
  "abcd": {
    "start": "./start.sh",
    "stop": "./stop.sh",
    "status": "./status.sh"
  }
}

Then, you will need to provide those hook scripts as part of your app.

Please be sure to chmod +x *.sh so that your hook scripts are executable.

start.sh

Following is an example for the start script. It submits a file named main (should be provided by each app) through qsub. It stores jobid so that we can monitor the job status.

1
2
3
4
5
6
#!/bin/bash

#return code 0 = job started successfully.
#return code non-0 = job failed to start

qsub -d $PWD -V -o \$PBS_JOBID.log -e \$PBS_JOBID.err main > jobid

stop.sh

Following is an example for the stop script. This script reads the jobid created by the start script and calls qdel to stop it.

1
2
#!/bin/bash
qdel `cat jobid`

status.sh

The status hook is a bit more complicated. It needs to return various exit codes based on the status of the app. It uses the jobid stored by start script to query the job status with qstat PBS command.

Anything you output to stdout will be used to set the task's status message. For example, you can output the last line from the log file to relay the last log entry to the users on Brainlife.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
#!/bin/bash

#return code 0 = running
#return code 1 = finished successfully
#return code 2 = failed
#return code 3 = unknown (retry later)

if [ ! -f jobid ];then
    echo "no jobid - not yet submitted?"
    exit 1
fi

jobid=`cat jobid`
if [ -z $jobid ]; then
    echo "jobid is empty.. failed to submit?"
    exit 3
fi

jobstate=`qstat -f $jobid | grep job_state | cut -b17`
if [ -z $jobstate ]; then
    echo "Job removed before completing - maybe timed out?"
    exit 2
fi

case "$jobstate" in
Q)
    showstart $jobid | grep start
    exit 0
    ;;
R)
    #get last line of last log touched
    logfile=$(ls -rt *.log | tail -1)
    tail -1 $logfile
    exit 0
    ;;
H)
    echo "Job held.. waiting"
    exit 0
    ;;
C)
    exit_status=`qstat -f $jobid | grep exit_status | cut -d'=' -f2 | xargs`
    if [ $exit_status -eq 0 ]; then
        echo "finished with code 0"
        exit 1
    else
        echo "finished with code $exit_status"
        exit 2
    fi
    ;;
*)
    echo "unknown job status $jobstate .. will check later"
    exit 3
    ;;

esac

Comments