diff --git a/bih-cluster/docs/best-practice/software-installation-with-conda.md b/bih-cluster/docs/best-practice/software-installation-with-conda.md index 273c9c648..d993661d6 100644 --- a/bih-cluster/docs/best-practice/software-installation-with-conda.md +++ b/bih-cluster/docs/best-practice/software-installation-with-conda.md @@ -10,7 +10,10 @@ to do so. We will provide you with a description on how to install conda and how to use it. Of course there are many online resources that you can also use. Please find a list at the end of the document. -Also note that some system-level software is managed through environment modules. +!!! Warning + Following a change in their terms of service Anaconda Inc. has started to demand payment from research institutions + for using both Anaconda, Miniconda, and the defaults channel. As a consequence, + usage of this software is prohibited and we're recommending the alternative free "miniforge" distribution instead. ## Premise When you logged into the cluster, please make sure that you also executed `srun` to log into a computation node and perform the software installation there. @@ -19,24 +22,21 @@ When you logged into the cluster, please make sure that you also executed `srun` ```bash hpc-login-1:~$ srun --mem=5G --pty bash -i -hpc-cpu-123:~$ wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -hpc-cpu-123:~$ bash Miniconda3-latest-Linux-x86_64.sh -b -f -p $HOME/work/miniconda -hpc-cpu-123:~$ eval "$(/$HOME/work/miniconda/bin/conda shell.bash hook)" +hpc-cpu-123:~$ wget https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-Linux-x86_64.sh +hpc-cpu-123:~$ bash Miniforge3-Linux-x86_64.sh -b -f -p $HOME/work/miniforge +hpc-cpu-123:~$ eval "$(/$HOME/work/miniforge/bin/conda shell.bash hook)" hpc-cpu-123:~$ conda init hpc-cpu-123:~$ conda config --set auto_activate_base false ``` -This will install conda to `$HOME/work/miniconda`. +This will install conda to `$HOME/work/miniforge`. You can change the path to your liking, but please note that your `$HOME` folder has limited space. The `work` subfolder however has a bigger quota. More about this [here](../storage/home-quota.md). -To make bioinformatics software available, we have to add the `bioconda` and -some other channels to the conda configuration: +To make bioinformatics software available, we have to add the `bioconda` channel to the conda configuration: ```bash hpc-cpu-123:~$ conda config --add channels bioconda -hpc-cpu-123:~$ conda config --add channels default -hpc-cpu-123:~$ conda config --add channels conda-forge ``` ## Installing software with conda diff --git a/bih-cluster/docs/how-to/connect/gpu-nodes.md b/bih-cluster/docs/how-to/connect/gpu-nodes.md index 23d9c36ec..abf9e6d18 100644 --- a/bih-cluster/docs/how-to/connect/gpu-nodes.md +++ b/bih-cluster/docs/how-to/connect/gpu-nodes.md @@ -31,14 +31,14 @@ Your job will be automatically placed in the Slurm `gpu` partition and allocated ## Usage example ### Preparation -We will setup a miniconda installation with `pytorch` testing the GPU. +We will setup a miniforge installation with `pytorch` testing the GPU. If you already have this setup then you can skip this step ```bash hpc-login-1:~$ srun --pty bash -hpc-cpu-1:~$ wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -hpc-cpu-1:~$ bash Miniconda3-latest-Linux-x86_64.sh -b -p ~/work/miniconda3 -hpc-cpu-1:~$ source ~/work/miniconda3/bin/activate +hpc-cpu-1:~$ wget https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-Linux-x86_64.sh +hpc-cpu-1:~$ bash Miniforge3-Linux-x86_64.sh -b -p ~/work/miniforge +hpc-cpu-1:~$ source ~/work/miniforge/bin/activate hpc-cpu-1:~$ conda create -y -n gpu-test pytorch cudatoolkit=10.2 -c pytorch hpc-cpu-1:~$ conda activate gpu-test hpc-cpu-1:~$ python -c 'import torch; print(torch.cuda.is_available())' @@ -80,7 +80,7 @@ nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2019 NVIDIA Corporation Built on Wed_Oct_23_19:24:38_PDT_2019 Cuda compilation tools, release 10.2, V10.2.89 -hpc-gpu-1:~$ source ~/work/miniconda3/bin/activate +hpc-gpu-1:~$ source ~/work/miniforge/bin/activate hpc-gpu-1:~$ conda activate gpu-test hpc-gpu-1:~$ python -c 'import torch; print(torch.cuda.is_available())' True diff --git a/bih-cluster/docs/how-to/software/tensorflow.md b/bih-cluster/docs/how-to/software/tensorflow.md index fd276a55c..f884f7463 100644 --- a/bih-cluster/docs/how-to/software/tensorflow.md +++ b/bih-cluster/docs/how-to/software/tensorflow.md @@ -155,7 +155,7 @@ print(tf.nn.softmax(predictions).numpy()) #SBATCH --partition=gpu #SBATCH --gres=gpu:tesla:1 -source $HOME/work/miniconda3/bin/activate +source $HOME/work/miniforge/bin/activate conda activate python-tf python tf_script.py &>tf-out.txt diff --git a/bih-cluster/docs/ondemand/interactive.md b/bih-cluster/docs/ondemand/interactive.md index 21ba84ba3..29bf431c6 100644 --- a/bih-cluster/docs/ondemand/interactive.md +++ b/bih-cluster/docs/ondemand/interactive.md @@ -167,7 +167,7 @@ dependencies: -Starting the Rstudio session via the OnDemand portal works almost as described above (see Example 1). However, you do have to select \`miniconda\` as R source and provide the path to your miniconda installation and (separated by a colon) the name of the (newly created) conda enviroment you want to use. +Starting the Rstudio session via the OnDemand portal works almost as described above (see Example 1). However, you do have to select \`miniforge\` as R source and provide the path to your miniforge installation and (separated by a colon) the name of the (newly created) conda enviroment you want to use. ![](figures/ondemand-interactive-custom-conda.png){: style="width:90%;" .center} diff --git a/bih-cluster/docs/overview/storage.md b/bih-cluster/docs/overview/storage.md index ed0bc8f53..d164808ed 100644 --- a/bih-cluster/docs/overview/storage.md +++ b/bih-cluster/docs/overview/storage.md @@ -1,79 +1,79 @@ -# Nodes and Storage Volumes - -!!! info "No mounting **on** the cluster itself." - - For various technical and security-related reasons it is not possible to mount anything on the cluster nodes by users. - For mounting the cluster storage on your computer, please read [Connecting: SSHFS Mounts](../connecting/advanced-ssh/linux.md#file-system-mount-via-sshfs). - -This document gives an overview of the nodes and volumes on the cluster. - -## Cluster Layout - -![](figures/Cluster_Layout.png) - -## Cluster Nodes - -The following groups of nodes are available to cluster users. -There are a number of nodes that are invisible to non-admin staff, hosting the queue master and monitoring tools and providing backup storage for key critical data, but these are not shown here. - -- `hpc-login-{1,2}` - - available as `hpc-login-{1,2}.cubi.bihealth.org` - - do not perform any computation on these nodes! - - each process may at most use 1 GB of RAM -- `med0101..0124,0127` - - 25 standard nodes - - Intel Xeon E5-2650 v2 @2.60Ghz, 16 cores x2 threading - - 128 GB RAM -- `med0133..0164` - - 32 standard nodes - - Intel Xeon E5-2667 v4 @3.20GHz, 16 cores x 2 threading - - 192 GB RAM -- `med0201..0264` - - 64 nodes with Infiniband interconnect - - Intel Xeon E5-2650 v2 @2.60Ghz, 16 cores x2 threading - - 128 GB RAM -- `med0301..0304` - - 4 nodes with 4 Tesla V100 GPUs each -- `med0401..0405` special purpose/high-memory machines - - Intel Xeon E5-4650 v2 @2.40GHz, 40 cores x2 threading - - `med0401` and `med0402` - - 1 TB RAM - - `med0403` and `med0404` - - 500 GB RAM - - `med0405` - - 2x "Tesla K20Xm" GPU accelleration cards (cluster resource `gpu`) - - access limited to explicit GPU users -- `med0601..0616` - - 16 nodes owned by CUBI - - Intel Xeon E5-2640 v3 @2.60Ghz - - 192 GB RAM -- `med0618..0633` - - 16 nodes owned by CUBI - - Intel Xeon E5-2667 v4 @3.20GHz, 16 cores x 2 threading - - 192 GB RAM -- `med0701..0764` - - 64 standard nodes - - Intel Xeon E5-2667 v4 @3.20GHz, 16 cores x 2 threading - - 192 GB RAM - -## Cluster Volumes and Locations - -The cluster has 2.1 PB of legacy fast storage, currently available at `/fast`, as well as 1.6 PB of next-generation fast storage, available at `/data/cephfs-1`. -Additionally 7.4 PB of slower "Tier 2" storage is available at `/data/cephfs-2`. -Storage is provided by a Ceph storage cluster and designed for massively parallel access from an HPC system. -In contrast to "single server" NFS systems, the system can provide large bandwidth to all cluster nodes in parallel as long as large data means relatively "few" files are read and written. - -Storage is split into three sections: - -- `home` -- *small, persistent, and safe storage*, e.g., for documents and configuration files (default quota of 1 GB). -- `work` -- *larger and persistent storage*, e.g., for your large data files (default quota of 1 TB). -- `scratch` -- *large and non-persistent storage*, e.g., for temporary files, files are automatically deleted after 2 weeks (default quota of 10 TB; deletion not implemented yet).) - -Each user, group, and project has one or more of these sections each, e. g. for users: - -- `/data/cephfs-1/home/users/$NAME` -- `/data/cephfs-1/home/users/$NAME/work` -- `/data/cephfs-1/home/users/$USER/scratch` - - -See [Storage and Volumes: Locations](../storage/storage-locations.md) for more informatin. +# Nodes and Storage Volumes + +!!! info "No mounting **on** the cluster itself." + + For various technical and security-related reasons it is not possible to mount anything on the cluster nodes by users. + For mounting the cluster storage on your computer, please read [Connecting: SSHFS Mounts](../connecting/advanced-ssh/linux.md#file-system-mount-via-sshfs). + +This document gives an overview of the nodes and volumes on the cluster. + +## Cluster Layout + +![](figures/Cluster_Layout.png) + +## Cluster Nodes + +The following groups of nodes are available to cluster users. +There are a number of nodes that are invisible to non-admin staff, hosting the queue master and monitoring tools and providing backup storage for key critical data, but these are not shown here. + +- `hpc-login-{1,2}` + - available as `hpc-login-{1,2}.cubi.bihealth.org` + - do not perform any computation on these nodes! + - each process may at most use 1 GB of RAM +- `med0101..0124,0127` + - 25 standard nodes + - Intel Xeon E5-2650 v2 @2.60Ghz, 16 cores x2 threading + - 128 GB RAM +- `med0133..0164` + - 32 standard nodes + - Intel Xeon E5-2667 v4 @3.20GHz, 16 cores x 2 threading + - 192 GB RAM +- `med0201..0264` + - 64 nodes with Infiniband interconnect + - Intel Xeon E5-2650 v2 @2.60Ghz, 16 cores x2 threading + - 128 GB RAM +- `med0301..0304` + - 4 nodes with 4 Tesla V100 GPUs each +- `med0401..0405` special purpose/high-memory machines + - Intel Xeon E5-4650 v2 @2.40GHz, 40 cores x2 threading + - `med0401` and `med0402` + - 1 TB RAM + - `med0403` and `med0404` + - 500 GB RAM + - `med0405` + - 2x "Tesla K20Xm" GPU accelleration cards (cluster resource `gpu`) + - access limited to explicit GPU users +- `med0601..0616` + - 16 nodes owned by CUBI + - Intel Xeon E5-2640 v3 @2.60Ghz + - 192 GB RAM +- `med0618..0633` + - 16 nodes owned by CUBI + - Intel Xeon E5-2667 v4 @3.20GHz, 16 cores x 2 threading + - 192 GB RAM +- `med0701..0764` + - 64 standard nodes + - Intel Xeon E5-2667 v4 @3.20GHz, 16 cores x 2 threading + - 192 GB RAM + +## Cluster Volumes and Locations + +The cluster has 2.1 PB of legacy fast storage, currently available at `/fast`, as well as 1.6 PB of next-generation fast storage, available at `/data/cephfs-1`. +Additionally 7.4 PB of slower "Tier 2" storage is available at `/data/cephfs-2`. +Storage is provided by a Ceph storage cluster and designed for massively parallel access from an HPC system. +In contrast to "single server" NFS systems, the system can provide large bandwidth to all cluster nodes in parallel as long as large data means relatively "few" files are read and written. + +Storage is split into three sections: + +- `home` -- *small, persistent, and safe storage*, e.g., for documents and configuration files (default quota of 1 GB). +- `work` -- *larger and persistent storage*, e.g., for your large data files (default quota of 1 TB). +- `scratch` -- *large and non-persistent storage*, e.g., for temporary files, files are automatically deleted after 2 weeks (default quota of 10 TB; deletion not implemented yet).) + +Each user, group, and project has one or more of these sections each, e. g. for users: + +- `/data/cephfs-1/home/users/$NAME` +- `/data/cephfs-1/home/users/$NAME/work` +- `/data/cephfs-1/home/users/$USER/scratch` + + +See [Storage and Volumes: Locations](../storage/storage-locations.md) for more informatin. diff --git a/bih-cluster/docs/slurm/snakemake.md b/bih-cluster/docs/slurm/snakemake.md index d577f730a..25ac5bbb4 100644 --- a/bih-cluster/docs/slurm/snakemake.md +++ b/bih-cluster/docs/slurm/snakemake.md @@ -4,8 +4,8 @@ This page describes how to use Snakemake with Slurm. ## Prerequisites -- This assumes that you have Miniconda properly setup with Bioconda. -- Also it assumes that you have already activated the Miniconda base environment with `source miniconda/bin/activate`. +- This assumes that you have Miniforge properly setup with Bioconda. +- Also it assumes that you have already activated the Miniforge base environment with `source miniforge/bin/activate`. ## Environment Setup