Troubleshooting
Note
This is a starting list seeded from the error and warning messages CUNQA currently emits. It is meant to grow over time as new issues and resolutions are collected.
Deployment (qraise)
- “qraise needs two mandatory arguments”
Both the number of vQPUs (
-n/--num-qpus) and the maximum time (-t/--time) must be provided. The time follows the SLURM format[D-]HH:MM:SS.- “Simulator <name> is not available for <scheme> communications simulation”
The chosen
--simulatordoes not support the requested communication scheme. Check the supported combinations in Simulators. Note also that the value passed to--simulatoris the CUNQA name (e.g.Munich), not the library name (MQT-DDSIM).- “There are QPUs with the same family name as the provided”
A group of vQPUs with that
--familyname is already deployed. Choose a different family name, or release the existing one withqdrop --family <name>first.- “At this moment, only Aer supports GPU simulation”
GPU execution (
--gpu/--gpu-name) is only available with--simulator Aer, and requires a GPU-enabled build (-DUSE_GPU=ON); see Installation.- “cores_per_qpu must be equal or greater than 2”
The quantum-communications scheme reserves one core per vQPU for the shared simulator process, so
--cores-per-qpumust be at least2.- “sbatch submission failed” (Python
qraise) The underlying
sbatchcall returned an error. Inspect the reportedstderr— common causes are an invalid partition, requesting more resources than available, or a malformed time string.
Querying / releasing (qinfo / qdrop)
- “Could not open the QPUs info file” / “There are not deployed QPUs!”
No vQPUs are currently registered. Deploy some with
qraisefirst. This file lives under$STORE/.cunqa/qpus.json.- “Problem accessing to the QPUs on the current node. Probably the command was run on a login node”
qinfo --mynoderelies on theSLURMD_NODENAMEvariable, which is only set on compute nodes. Run it from within an allocation, or query a specific node withqinfo <node>.- “No qraise jobs are currently running with the specified id / family names”
The given IDs or family names did not match any deployed vQPU. Use
qinfoto list what is currently deployed.- “You must specify either the IDs or the family name (with –family) … or use the –all flag”
qdropneeds exactly one target selector: job IDs,--family, or--all. Combining IDs with--familyis not allowed.
Execution (run)
- “No QPUs were provided”
runwas called withqpus=None. This usually meansget_QPUsreturned nothing — see the next entries.- “There are not enough QPUs”
More circuits than vQPUs were submitted. Provide at least as many vQPUs as circuits (a single circuit may be broadcast to several vQPUs, but not the other way around).
- “Not enough data qubits / comm qubits in the QPU for the circuit”
The circuit needs more (data or communication) qubits than the target vQPU’s backend provides. Reduce the circuit size or deploy vQPUs with a larger backend.
- “You are searching for QPUs in a login node, none are found” (warning)
get_QPUswas called withoutco_located=Truefrom a login node. Useco_located=Trueif the vQPUs were deployed in co-located mode, or run from a compute node.- “No QPUs were found” / “No QPUs were found with the characteristics provided” (warning)
No registered vQPU matched the requested
co_located/familyfilters. Verify the family name and the access mode used at deployment withqinfo.