Content
- Check the status of submitted jobs
- Delete a job from the queue
- Hold back/delay a job from executing
- Release previously held jobs
- Get information about completed jobs
Check the status of submitted jobs
To query the status of submitted batch jobs, execute:
scc % qstat -u userID
job-ID prior name user state submit/start at queue slots ja-task-ID
-----------------------------------------------------------------------------------------------------------------
1000001 0.10000 network_gr userID r 05/19/2014 09:17:53 b@scc-bd7.scc.bu.edu 1
1000002 0.00000 network_gr userID qw 05/19/2014 09:14:53 16
| Understanding qstat output | |
| Queue Column | Description |
|---|---|
| job-ID | Lists the job ID. All jobs have a unique job ID that can be used to reference a particular job. |
| prior | Priority of the job. This number is set by the SCC batch scheduling system and its value is set based on a “fair share” system to make sure all researches have reasonably equal access to the SCC resources. |
| name | Name of the job. A user can specify the name of the job using the -N option to the qsub command. Otherwise the system will use the batch script (or binary executable) name as a job’s name. |
| userID | User SCC login ID. |
| state | The state of the job: (r) – running; (qw) – waiting to run; (hqw) – on hold, waiting to run; (Eqw) – job in error state; (s) – suspended; (t) – transfering. |
| submit/start | Time when the job was submitted. When the job is running, this field is updated with the time the job started. |
| queue | The queue name and the node ID on which the job is running. |
| slots | The number of slots the job requested. |
| task | The task ID if the job was submitted as a job array. For all other types of jobs this field is empty. |
Delete a job from the queue
scc % qdel jobID
userID has deleted job jobID Hold back/delay a job from executing
scc % qhold jobID
modified hold of job jobID Release previously held jobs
scc % qrls jobID
modified hold of job jobID Get information about completed jobs
The qacct tool is an SGE tool to report on usage over the entire SCC cluster. If the command is executed without arguments it will output the aggregate total of SCC usage. For the user or a project, a more useful output could be obtained using various options to the qacct command. For example, to retrieve the information about completed jobs:
scc % qacct -j jobID
To list the full statistics for all the jobs a user ran during the past week, he can execute:
scc % qacct -d 7 -o userID -j For every job that ran during the specified period of time, this command will list the exit status, user, system and wallclock time, memory usage, queue name, hostname, and some other relevant information. Read the man page for qacct for the list of all available options and their usage.