Learning Forum

[kevin1986]

Hi guys,

We have a 4 node cluster with amd epyc CPUs. They are connected by mellanox switch. However, sometimes I found some node would disconnect with the others, either this node cannot ssh to the others, or this node's folder cannot be mounted. It happens quite randomly. I tried different version of Fluent, tried different mellanox switch, tried different DAC cables, tried different centOS, tried intelmpi and openmpi, it still has this problem.

I install Fluent in a shared folder, all nodes can access this folder. Do I need to install a local copy of fluent on each node?

I really dont know why. Any experts can give any clue?

Thanks.

[RK]

Hello Kevin,
You might want to check with your IT department on this first.

[kevin1986]

Hi RK Thanks for your reply. Actually I work for IT department... I just have a question. When we build a cluster, we have two methods:
1) we install ansys on each node.
2) we install ansys in a folder, then share this folder by mount -o tcp,port=2049
then each node can run Ansys.
Right now we use method 2). I also found method 1) works. I am not sure if method 1) is preferable when builing a cluster.
Thanks.

[kevin1986]

These days, we tried different mellanox switch, different cables, different IB cards, we still have this problem. We are running Fluent 2022 linux version. I appreciate any kind of help or advices :-)

[RK]

Hi Kevin,
We acknowledge your response. One of us will get back to you shortly.