Closed zhangxiaoshuotttt closed 3 months ago
Thanks for your question @zhangxiaoshuotttt, in order to run on a specific cluster, you will need to write your own sub-class that inherits from Slurm but provides some additional information like partition
. Please see, for example, the Chinook sub-system (https://github.com/adjtomo/seisflows/blob/master/seisflows/system/chinook.py), which represents a sort of minimum working example of what you'll need to write.
You can copy-paste and rename this class for your own purpose, and make adjustments to the internal _partitions
variable to match the cores-per-node and partition names on your system (https://github.com/adjtomo/seisflows/blob/master/seisflows/system/chinook.py#L50-L52). Similarly the run_call_header
and submit_call_header
will need to be modified to match how you would call SLURM directives on your system.
Let me know if that makes sense to you! Happy to help you workshop this and to later include your subclass in the main package.
Thank you for your advice. I agree and will try it now. I'll share my experience if it works and hope it helps others.
I apologize for the delayed response. After testing on our school's cluster, I have found that the method suggested by @bch0w works well. For beginners who wish to use Seisflows on a cluster managed by Slurm, all that is required is to create a sub-class that inherits from Slurm and provides additional information. @bch0w has explained this clearly. Thanks again for this advice.
Great to hear @zhangxiaoshuotttt! Thanks for following up on this, glad that worked for you. If you are interested in contributing your sub class to the package you are more than welcome to submit a Pull Request to the devel
branch. I'll close this as complete for now but if you run into any issues please feel free to reopen or start a new one.
Hello @bch0w ,
I am encountering issues while using Seisflows on the CentOS cluster, which is managed by
Slurm
.Initially, I tried setting up the system by implementing process
seisflows par system slurm
, but I encountered some problems.I tried to identify the root cause by going through the codes, and it might be because I didn't set up _partitions. I attempted to set up partition and node_size, but was unsuccessful.
Therefore, I changed the system to
cluster
, which worked well. However, I am now facing a different issue.I have provided all the information that might be relevant to help you understand the situation better.
Thank you for your time and assistance. All my settings
Best regards, Zhang