# Install / Conda

Conda is an awesome package manager, but was slow and used a ton of memory as dependencies increased. Recent (23.10+) versions of Conda now use the libmamba solver, the super-fast and super-lightweight solver from Mamba. If you're experiencing a suspiciously slow Harpy installation, either update Conda to at least version 23.10 or use Mamba/Pixi. If conda is forbidden on your system, see below.

Sometimes conda channel priorities may give you grief, be it when trying to install Harpy, or Harpy trying to install a particular workflow's dependencies. Although Conda recommends setting channel priorities to strict, we find that it sometimes causes the solver (within or without Snakemake) to fail to find the necessary software. One typical way to circumvent that is to set the channel priority to something other than strict:

conda config --set channel_priority true
# or #
conda config --set channel_priority false

Unless using --container, Harpy leans on Snakemake to install small conda environments necessary to accomplish the tasks in workflows. The most common issue we tend to see is errors during this process.

It's tempting to install software into the base environment conda/mamba ships with, but that behavior is discouraged (by Conda itself, not just us). Modificatons to base can mess with all sorts of things in other environments (see this issue), so we recommend installing Harpy (or anything) into literally any other environment. If you find conda things just aren't working right and you know you have at some point modified the base environment, you can reset your base environment. It's not guaranteed, but doing that might just fix things:

conda install --rev 0 --name base

Some HPC configurations prohibit the use of conda. If that's preventing you from installing Harpy, then you can try to install it as a python package using pip. You will still need to manually install the package dependencies provided below. Since conda is disabled, you will need to use the --container option for most of the commands.

A not-uncommon HPC setup is to have the login node connected to the internet, but the worker nodes have no internet access. This obviously creates a problem if Snakemake needs to download software dependencies (conda or container). To facilitate this kind of setup, Harpy has the deps command, which runs a fake workflow to download and install workflow dependencies. The deps conda and deps container commands are meant to be used on the internet-connected login node in circumstances like this.

Create the conda environments required by Harpy's workflows (e.g. phase). Only useful for specific HPC configurations where worker nodes do not have internet access to let snakemake install conda packages itself.

harpy deps conda workflows...

harpy deps conda all

You can provide any combination of workflow names to selectively install only the necessary environments, or all to install (you guessed it!) all of them:

• all
• align
• assembly
• metassembly
• phase
• qc
• r
• simulations
• stitch
• variants

Manually pull the Harpy dependency container from dockerhub and convert it into an Apptainer .sif file. Only useful for specific HPC configurations where worker nodes do not have internet access to let snakemake download the container itself. Run this command again without arguments to use it.

harpy deps container