Infiniband interface down on Bamboo

HPC Support HPC issues
1

Hello all,

I’m trying to run a dask scheduler/worker system on bamboo. I usually use infiniband (ib0) on both the scheduler node and the worker nodes. However in this case I was getting timeout when the workers attempted to contact the scheduler.

Switching to ethernet (eno1) was enough to resolve the problem.

Is there a current infiniband issue on Bamboo?

Regards,
Berk

2

Dear Berk,

Actually IB is working fine on bamboo we have done a lot of bench.
As beegfs is using rdma, we can assume that all seems to be ok.

Could you please share your config files to check?

Best regards,