Add an FAQ entry about a certain GPU-aware MPI error (#4811)

jrood-nrel · web-flow · commit ef7ca320dc83 · 2025-11-20T11:00:10.000-08:00
diff --git a/Docs/sphinx_documentation/source/Faq.rst b/Docs/sphinx_documentation/source/Faq.rst
@@ -172,6 +172,13 @@ to the device and pass a device pointer function object `DevicePtrIF` into the `
 |
 |
 
+**Q.** I'm getting errors when running with GPU-aware MPI
+
+**A.** While other problems may exist. One thing to check, if the machine is using Slurm, is the `cgroup.conf` file. If it contains `ConstrainDevices=yes`, then IPC can be impacted, which means bindings such as `--gpu-bind=closest` should not be used. Instead try `--gpu-bind=none`.
+
+.. _`This Slurm issue provides more information`: https://support.schedmd.com/show_bug.cgi?id=17875
+
+
 More Questions
 --------------
 

Original file line number	Diff line number	Diff line change
@@ -172,6 +172,13 @@ to the device and pass a device pointer function object `DevicePtrIF` into the `
`172`	`172`	`\|`
`173`	`173`	`\|`
`174`	`174`
	`175`	`+Q. I'm getting errors when running with GPU-aware MPI`
	`176`	`+`
	`177`	+A. While other problems may exist. One thing to check, if the machine is using Slurm, is the `cgroup.conf` file. If it contains `ConstrainDevices=yes`, then IPC can be impacted, which means bindings such as `--gpu-bind=closest` should not be used. Instead try `--gpu-bind=none`.
	`178`	`+`
	`179`	+.. _`This Slurm issue provides more information`: https://support.schedmd.com/show_bug.cgi?id=17875
	`180`	`+`
	`181`	`+`
`175`	`182`	`More Questions`
`176`	`183`	`--------------`
`177`	`184`