Error loading shared libraries¶
Most things in
Salvus are statically linked, except MPI, thus dynamic MPI
libraries must be available. If these are not present, errors like the
error while loading shared libraries: libmpicxx.so.12: cannot open shared object file: No such file or directory
On sites without central job submission system (thus usually single nodes),
just let the Mondaic downloader also download MPI. It will but the MPI
libraries in the
lib/ folder and
SalvusFlow will set the
accordingly when running Salvus.
For other sites, please load a suitable MPI module (see
here for more details). In some cases the
module system might not set the
$LD_LIBRARY_PATH - then it has to be done
manually in the
SalvusFlow config file for the corresponding site.
Jupyter kernel dies¶
If you see the Jupyter notebook kernel dying when importing parts of Salvus make sure that you have a valid license and an active internet connection. If one of these is the issue, there should be an error message in the terminal that started Jupyter notebook. For technical reasons we cannot show this message in the Jupyter notebook.
Certificate errors while running the downloader¶
On old systems (or just having an old version of
curl which might for example
also happen when inside a
conda environment) the certificate might not be
valid, so errors like this one might occur when attempting to run Mondaic's
curl: (77) error setting certificate verify locations: CAfile: /etc/pki/tls/certs/ca-bundle.crt CApath: none
In these cases you could attempt to fix your
curl installation or just
manually do what the
get.mondaic.com script is doing:
Download either https://get.mondaic.com/salvus_downloader_linux or https://get.mondaic.com/salvus_downloader_osx (depending on your operating system) by whatever means possible (
browser, copy on other machine and
scpit over, ...).
Give it execution rights and run it:
shell $ chmod u+x salvus_downloader_* $ ./salvus_downloader_*
The rest is business as usual.
An error like
occurs if attempting to run a binary that has instructions that the current CPU cannot interpret. If you see this when running Salvus you likely downloaded the wrong binary for your system. Please see our documentation about instruction sets for more details.
Our downloader should auto-detect your CPU and choose the correct binaries. If that detection is faulty, please let us know.
Error during import of salvus_flow¶
There was a recent update of the
arrow package, which is used by SalvusFlow.
For arrow versions prior to
0.14.4 you might receive the following error:
AttributeError: module 'arrow.factory' has no attribute 'ArrowParseWarning'
To check which version you have installed, run the following line in a terminal
python -c "import arrow; print(arrow.__version__)"
To upgrade your local installation of arrow, run
pip install --upgrade arrow