I was trying to run a workflow on CERN’s REANA installation using the HTCondor backend. The instructions for auto-config of Rucio in REANA shown at Rucio - Documentation only work with the Kubernetes backend, right?
Hello @atanasi ,
I confirm that Rucio support only works with the Kubernetes backend. I will update the docs to make this clear.
Would you be interested in REANA supporting Rucio on HTCondor? If so, I can discuss this with the team to see what we can do about it.
Hi @madonado,
Well, in principle yes, I am interested in REANA supporting Rucio in the HTCondor backend. I can get around the limitation by uploading my proxy to the REANA workspace and my own Rucio cfg file, but having the same Rucio support as in the Kubernetes case would be convenient.
I am from the Einstein Telescope community. We do not have our own installation of REANA yet, but we are planning to have one. It is not clear if we will use an HTCondor cluster, a Slurm cluster, a Kubernetes cluster, or more than one of them. So far we are only testing running on HTCondor backend (in the REANA CERN installation) a workflow that we already ran on Kubernetes backend. We are doing this, because our users run their analyses on HTCondor clusters. And here I have another question for you. The rewritting of an HTCondor DAG to one of the REANA-supported workflow languages is (almost) the same independedently of the used backend, so the choice of backend would depend solely on what kind of cluster is our REANA installation attached to. Is that correct? To be more specific, REANA doesn’t support for example giving it a condor submit file for each workflow step instead of specifying the commands to run. If it would support that, then I think we would most likely choose to use the HTCondor backend (because it would be easier for us to translate an HTCondor DAG to run on REANA with HTCondor backend), and then I would say to you that yes, we definitely need to have the Rucio support with HTCondor.