Skip to content

Latest commit

 

History

History
102 lines (84 loc) · 3.86 KB

README.md

File metadata and controls

102 lines (84 loc) · 3.86 KB
Raw

ccutils

Various Slurm utilities scripts for Compute Canada staff and users.

Installation

Run the following to install in ~/.local/bin/ by default.

bash install.sh

Install in a different location with :

PREFIX=~/bin bash install.sh

Feel free to then rename the scripts to your likings.

Contribute

  1. Add your script
  2. Document its usage

Scripts

recent_usage

Scripts to display recent core usage of an account. By default, it display 14 days.

$ recent_usage.sh def-account [days]
In the past 14 days, project def-account has used on average 0 cores and 0 GPUs

squeue_prio

Script to display top 10 jobs and their current priority.

$ squeue_prio.sh  <userX>
          JOBID   USER  PRIORITY
       32402825  userX  2519571
       32440131  userX  2318474
       32149496  userX  2300399
       32438832  userX  2224271
       32438835  userX  2224271
       32438838  userX  2224271
       32438839  userX  2224271
       32438843  userX  2224271
       32438844  userX  2224271

partition-queue

Script to show a numbered list of jobs waiting in a particular partition or list all partitions.

$ partition-queue -h
Usage:
partition-queue <partition_name>[,<partition_name>[...]] # show numbered list of jobs waiting in <partition_name>
partition-queue [-p|-l]                                  # show list of all partitions
partition-queue [-h|--help]                              # show this help

$ partition-queue -l
available partitions:
cpubase_interac
cpubase_bynode_b1
cpubase_bynode_b2
cpubase_bynode_b3
cpubase_bynode_b4
cpubase_bynode_b5
cpubase_bynode_b6
...

$ partition-queue cpubase_bycore_b2 | head
     0	   JOBID     USER      ACCOUNT           NAME  ST  TIME_LEFT NODES CPUS       GRES MIN_MEM NODELIST (REASON) 
     1	12345671 user1234 def-someprof    some_job.sh  PD    7:59:00     1    1     (null)      2G  (Priority) 
     2	12345672 user1234 def-someprof    some_job.sh  PD    7:59:00     1    1     (null)      2G  (Priority) 
     3	12345673 user1234 def-someprof    some_job.sh  PD    7:59:00     1    1     (null)      2G  (Priority) 
     4	12345674 user1234 def-someprof    some_job.sh  PD    7:59:00     1    1     (null)      2G  (Priority) 
     5	12345675 user1234 def-someprof    some_job.sh  PD    7:59:00     1    1     (null)      2G  (Priority) 
     6	12345676 user1234 def-someprof    some_job.sh  PD    7:59:00     1    1     (null)      2G  (Priority) 
     7	12345677 user1234 def-someprof    some_job.sh  PD    7:59:00     1    1     (null)      2G  (Priority) 
     8	12345678 user1234 def-someprof    some_job.sh  PD    7:59:00     1    1     (null)      2G  (Priority) 
     9	12345679 user1234 def-someprof    some_job.sh  PD    7:59:00     1    1     (null)      2G  (Priority)

whenwilldrain.sh

Script takes a node-spec and prints sinfo and squeue information about those nodes so that one can see when the current jobs will finish.

$ whenwilldrain cluster[20-15]
PARTITION  AVAIL  TIMELIMIT  NODES  STATE NODELIST
general*      up 3-00:00:00      6  alloc cluster[20-25]

JOBID     USER       STATE       TIME_LEFT     NODELIST    END_TIME            
123401    user123    RUNNING      22:50:47    cluster24    2024-08-28T13:58:37    
123402    user123    RUNNING      22:50:47    cluster25    2024-08-28T13:58:37    
123402    user321    RUNNING    2-06:51:04    cluster20    2024-08-29T21:58:54    
123403    user321    RUNNING    2-06:51:04    cluster20    2024-08-29T21:58:54    
123404    user321    RUNNING    2-06:51:04    cluster21    2024-08-29T21:58:54    
123405    user321    RUNNING    2-06:51:04    cluster21    2024-08-29T21:58:54    
123406    user321    RUNNING    2-06:51:04    cluster22    2024-08-29T21:58:54    
123407    user321    RUNNING    2-06:51:04    cluster22    2024-08-29T21:58:54    
123408    user321    RUNNING    2-06:51:04    cluster23    2024-08-29T21:58:54