ARCC Commands

1 Core Hour Usage for User: chu_user | 2 Core Hour Usage for Project: chu_account | 3 Core Hour Usage: Differences to (current) Monthly Reports

This page details the use of a variety of ARCC specific cluster related commands.

Core Hour Usage for User: chu_user

Usage:

Calculate the total core usage hours for a user.
Syntax: chu_user [-u|y|m|Y|h]
options:
u <user>  : optional : username to look up - if not defined will use current user.
y <year>  : optional : year to look up - form YYYY e.g. 2023.
m <month> : optional : month to look up - form M e.g. 1 to 12.
                     : If no month defined will calculate for the entire year, or until 'now' if current year.
                     : If no date options define will calculate from beginning of current month.
Y         : optional : Calculate from beginning of current year to 'now' - overrides 'y' and 'm' if defined.
h         : optional : Print this help.
More details can be found at: https://arccwiki.atlassian.net/wiki/spaces/DOCUMENTAT/pages/1809907760/ARCC+Commands

Output:

[]$ chu_user -u salexan5
+===========================================================================================+
| Core Hour Usage: salexan5                                                                 |
| Interval: 2023-05-01T00:00:00 to 2023-05-24T13:45:35                                      |
| Hours from start of month: 589.76                                                         |
+-------------------------------------------------------------------------------------------+
| Non Investment Accounts:      |                             |                             |
| Breakdown by Account          | Total Jobs   Total CPUHs    | User Jobs    User CPUHs     |
| arccanetrain                  | 4348         1331.73        | 16           2.08           |
| matlabworkshop2022            | 54           215.09         | 0            0.0            |
| rongsongcodeoptim2022         | 18           29.89          | 0            0.0            |
+-------------------------------------------------------------------------------------------+
| Investment: inv-arcc          |                             |                             |
| Size (# of cpus): 132         |                             |                             |
| Across all accounts           | Total Jobs   Total CPUHs    | User Jobs    User CPUHs     |
|                               | 69           536.8          | 21           514.28         |
| Breakdown by Account          |                             |                             |
| arcc                          | 69           536.8          | 21           514.29         |
+===========================================================================================+

Label	Description

Label	Description
`Core Hour Usage: <username>`	The username the query is being made for. If not calling to check your own usage, you can only query users that you are associated to within your projects.
`Interval`	Starts the start to end date for the interval the query is running between.
`Hours from start of month`	Defines the number hours across the interval.
`Non Investment Accounts`	Lists the accounts that the user is part of, which are not associated with an investment. `\| Total Jobs Total CPUHs \| User Jobs User CPUHs \|` The first two columns show the “number of jobs” and the “CPU Hour usage” across the account for ALL users. The second two columns show the “number of jobs” and the “CPU Hour usage” across the account specifically for the user.
`Investment`	A user can be part of an account that is associated with an investment. Note: An investment can have multiple accounts associated with it.
`Size (# of cpus)`	The total number of cores across all the investment nodes.
`Across all accounts`	Same as `Non Investment Accounts` but details totals across ALL accounts within the investment.
`Breakdown by Account`	Same as `Non Investment Accounts` but details totals across individual accounts within the investment.

Note: A user can also be part of multiple investments. If this is the case there will be an ‘Investment’ section for each investment.

Core Hour Usage for Project: chu_account

Usage:

Output:

Label	Description

Label	Description
`Account Core Hour Usage: <project>`	The project (slurm account) the query is being made for. You can only query projects that yopu are part of.
`Interval`	Starts the start to end date for the interval the query is running between.
`Hours from start of month`	Defines the number hours across the interval.
If the account is part of an investment:	See Example 1
`Investment`	A user can be part of an account that is associated with an investment. Note: An investment can have multiple accounts associated with it.
`Size (# of cpus)`	The total number of cores across all the investment nodes.
`Days in Month: <days>`	Total number of days across the month/interval.
`Investment CPUHs: <cpu hours>`	Total CPU hours relating to the investment over the entire month/interval.
`Balance:`	The calculated as the “investment cpus + free amount - current number of CPU hours used”
`Across all accounts`	The columns show the “number of jobs” and the “CPU Hour usage” across ALL the accounts that are associated with the investment.
Account Details
<account>	The columns show the “number of jobs” and the “CPU Hour usage” across the account being queries.
By User
<username>	Breakdown of the total jobs/CPUs by user. If the `username` is a list of numbers, for example `10237419` then it means that user is no longer part of the cluster.

Core Hour Usage: Differences to (current) Monthly Reports

For calculating both user and account usage we are exploring an alternative approach to calculating usage compared to that currently implemented within the monthly reports.

Results do not include:

Jobs that were pre-empted.
Jobs that failed due to “Node Failure' - this is typically a hardware fail on the physical compute node - this is different to jobs that failed due to the application itself or experienced an out of memory issues (which is caused by the user not requesting enough memory for the job).
Job counts will typically be less as we do no include jobs that were cancelled while still pending i.e. the job had not been allocated any resources and had not started running.