Skip to content
Snippets Groups Projects
Commit b07fbab3 authored by Natasha Pavlovikj's avatar Natasha Pavlovikj
Browse files

Merge branch 'update_faq' into 'master'

Add highmem FAQ

See merge request !459
parents 669e598d 0dd6f3c4
No related branches found
No related tags found
1 merge request!459Add highmem FAQ
......@@ -29,6 +29,7 @@ weight = "95"
- [Can HCC provide training for my group?](#can-hcc-provide-training-for-my-group)
- [Can HCC provide help and resources for my workshop?](#can-hcc-provide-help-and-resources-for-my-workshop)
- [Where can I get training on using HCC resources?](#where-can-i-get-training-on-using-hcc-resources)
- [My job is submitted to the highmem partition and is pending with QOSMinMemory reason. What does this mean?](#my-job-is-submitted-to-the-highmem-partition-and-is-pending-with-qosminmemory-reason-what-does-this-mean)
---
......@@ -382,3 +383,25 @@ HCC provides free and low cost training events throughout the year. Most events
New events are posted on our [upcoming events page](https://hcc.unl.edu/upcoming-events) and announced through our [hcc-announce mailing list](https://hcc.unl.edu/subscribe-mailing-list).
Past events and their materials are also available on our [past events page](https://hcc.unl.edu/past-events).
#### My job is submitted to the highmem partition and is pending with QOSMinMemory reason. What does this mean?
The majority of nodes in the `batch` partition on Swan have 256GBs of RAM, with a few nodes with up to 2TBs of RAM. To ensure that the jobs that require lots of memory will run on the nodes with more RAM memory, SLURM uses the `highmem` partition, which is part of the `batch` partition. **This is not an actual partition, so it can not be separately used.** SLURM internally submits the job to both `highmem` and `batch` partitions, and depending on the requested RAM memory, allocates the requested resources. During this process, when checking the job status, you may see:
{{< highlight batch >}}
$ squeue -u demo
JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON)
1000000 highmem,b job_name demo PD 0:00 1 (QOSMinMemory)
{{< /highlight >}}
This message means that the job does not require high memory and it will be submitted to the `batch` partition when the requested resources are available. Once this internal process is completed, the `NODELIST(REASON)` message will be updated accordingly.
Please note that `highmem,b` is truncated from `highmem,batch`. The expanded output can be seen with:
{{< highlight batch >}}
$ squeue -u demo -o "%.18i %.20P %.8j %.8u %.2t %.10M %.6D %R"
JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON)
1000000 highmem,batch job_name demo PD 0:00 1 (QOSMinMemory)
{{< /highlight >}}
{{% notice note %}}
The number of nodes with high memory is limited, so please only request high amounts of memory if the job really needs it. Otherwise, you may encounter longer waiting times, lower submission priority and underutilized resources.
{{% /notice %}}
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment