Code of Conduct
By accepting the Terms and Conditions of Access, the user agrees to follow the Code of Conduct. Unlike the Terms and Conditions, we expect the Code to develop in response to the development of the service.
The HECToR service is shared by a large number of people and projects. We are sure that users will be keen to cooperate as colleagues to ensure that this common resource is run fairly and efficiently.
The Code of Conduct will be a set of rules of etiquette which we commend to users in order to achieve this end. If you follow them you can be reasonably sure that you are not being a nuisance to colleagues in other projects or interfering with the smooth running of the system.
We don't plan to administer the service with a heavy hand. In our experience, when users become aware that what they are doing is causing problems to others, they are anxious to change this. However, if you ignore the Code of Conduct we will bring it to your attention, and if you do so seriously or constantly, we will take it up with your Principal Investigator. And if necessary, we will hold jobs, or even kill them, in order to keep the system as a whole operating smoothly.
The Code of Conduct will change as the service evolves.
The HECToR Code of Conduct
Avoid filling the job queues
Under the current configuration users can have a maximum of 4 jobs running in any one queue at any one time. The maximum number of jobs a user can run in any one queue at any one time is restricted to 8. There is no system restriction on the number of jobs queued, however we would request that users queue no more than 8 jobs in any one queue at any one time. It is counterproductive to queue multiple jobs. It not only causes inconvenience for other users, but it can also increase the wait time on your own jobs due to the backfill scheduling that is implemented on HECToR . A number of techniques are available for running large numbers of jobs and these are documented in the HECToR Optimisation Guide If you do overfill the queues, we will request that jobs are removed.
Avoid running CPU intensive jobs on login nodes
The serial batch queues have been set up exactly for the purpose of enabling lengthy and heavily cpu-bound applications to be run efficiently without impacting on the wider user service. You can find documentation on using the serial queues here in the HECToR User Guide (opens in a new window/tab).
Avoid filling the /tmp partition on login nodes
Saturating /tmp causes disruption for other users. Please ensure that
you use your allocated home and work directories for any files. If you
require additional disk allocation please contact your Principal
Investigator who will be in a position to authorise this.
Avoid using HECToR as a route to other servers
Please do not piggyback from the login nodes to other servers.
