BSC HPC machines
Providentia’s download mode supports downloading GHOST and non-GHOST observational data, as well as model outputs, directly from BSC HPC systems.
In order to do this type of download, a BSC HPC account is required.
Supported HPC login nodes
Providentia can download GHOST and non-GHOST networks from the BSC HPC environment using available login nodes. By default, it connects to storage5 (or mn5 if storage5 is unavailable).
The login nodes used are defined in settings/dl_hpc.yaml:
transfer1.bsc.es(primary)transfer2.bsc.estransfer3.bsc.estransfer4.bsc.esglogin4.bsc.es
Downloads will attempt the nodes in the order listed. If the primary node is unavailable, the system automatically tries the next one in the list.
Download of networks from HPC machines
The network is essential to generate a report, as it provides access to the real observational data. At BSC, a large number of observational datasets are already stored on gpfs.
Download source paths:
Networks are saved under the
ghost_rootandnonghost_rootpaths specified in thelocalkey ofsettings/data_paths.yaml.Networks are retrieved from the
ghost_rootandnonghost_rootpaths specified in thestorage5ormn5key ofsettings/data_paths.yaml.
How to enable:
You must include at least one network in your configuration.
Set
dl_modetobothorobs(or answerboth/obsto the prompt “Which type of data do you want to download? Observational, modelled or both?”).For GHOST network downloads, answer
yto the prompt:
“Do you want to download observational data from the BSC remote machine? (Otherwise, GHOST observational data will be retrieved from Zenodo)”
or setdl_ghost_source = bscin your configuration.If the configuration file contains only network data and no models, network data will be downloaded automatically.
Download of interpolated models
An interpolated model consists of model output that have already been spatially interpolated to the observations. These models are ready to be used in modes such as report and dashboard. Most interpolated model data is stored on gpfs.
Download source paths:
Models are saved under the
mod_rootpath specified in thelocalkey ofsettings/data_paths.yaml.Models are retrieved from the
mod_rootpath specified in thestorage5ormn5key ofsettings/data_paths.yaml.
How to enable:
You must include at least one network and one model in your configuration.
Set
dl_modetobothormod(or answerboth/modto the prompt).Answer
yto the prompt:
“Model data was detected in the configuration file. Do you want to download the interpolated version? (Otherwise, the non-interpolated model data will be downloaded)”
or setdl_interpolated = Truein your configuration.
Download of non-interpolated models
Local non-interpolated downloads
Non-interpolated model data refers to model outputs that are ready to be interpolated against a network using interpolation mode. Most non-interpolated datasets are stored in esarchive, although some may already exist in gpfs.
Download source paths:
Models are saved under the
mod_to_interp_rootpath specified in thelocalkey insettings/data_paths.yaml.Models are retrieved from the paths specified in
settings/interp_models.yaml. To learn how to define models, please see the Defining models in interp_models.yaml section in Interpolation.If the model is not in the path, the system looks in the
mod_to_interp_rootpaths under thestorage5ormn5key insettings/data_paths.yaml.
How to enable:
You must include at least one model in your configuration.
Set
dl_modetobothormod(or answerboth/modto the prompt).Answer
nto the prompt:
“Model data was detected in the configuration file. Do you want to download the interpolated version? (Otherwise, the non-interpolated model data will be downloaded)”
or setdl_interpolated = Falsein your configuration.If the configuration file contains only model data and no networks, non-interpolated model data will be downloaded automatically.
HPC non-interpolated downloads
In some cases, model data exists in esarchive, but not all HPC machines have direct access to it. When interpolation needs to be performed on a machine without esarchive access, the model data must first be copied from esarchive to gpfs.
Download source paths:
Models are copied from
esarchiveto thegpfsmod_to_interp_rootfolder defined under thestorage5key insettings/data_paths.yaml.Only copies from paths specified in
settings/interp_models.yaml. To learn how to define models, please see the Defining models in interp_models.yaml section in Interpolation.
How to enable:
The download must be performed from the
storage5machine.
.env file
An .env file will appear in the Providentia root directory when using the download mode. It is designed to store specific user preferences.
PRV_USER: This setting specifies the username used to connect to the remote machines. It can be any valid username, e.g.:
bsc000000.PRV_PWD: This setting allows you to save the password needed for connecting to remote machines.
Note that the password is not required if you have configured a passwordless connection to the different servers.
Tutorial: SSH Key Autologon Only accessible for users with a BSC CAS account.
These values can be changed directly on the .env file and also be updated by Providentia during the next run.