Skip to content

Latest commit

 

History

History
96 lines (80 loc) · 4.11 KB

install.md

File metadata and controls

96 lines (80 loc) · 4.11 KB
Raw

Installation

Prerequisites

Before installing Cpipe, you need to make sure you have the following packages installed on your system. Most of these are very common packages and probably came pre-installed on your machine. However this is an exhaustive list:

Runtimes

  • Java 1.8 (preferably a build later than 1.8.0_20, since Cpipe has issues with this build)

Libraries

  • SQLite 3
  • zlib
  • bzip2
  • readline
  • ssl
  • xorg
  • curl
  • ncurses

Build Tools

  • gcc
  • gfortran
  • make

Utilities

  • curl
  • git

To see a complete and up-to-date list, have a look at the apt-get install command in the Dockerfile

If you're using a module system, you might have to module load some of these modules. If they are not available at all, they can be installed by yourself or a sysadmin with a command like sudo apt-get install -y libsqlite3-dev etc..

Swift Credentials

If you are installing Cpipe as part of the MGHA (Melbourne Genomics Health Alliance), you have access to the Cpipe asset bundle, which decreases the installation time to about an hour. We intend to make this bundle accessible to everyone in the future.

Download the swift credentials file in one of two ways. Either clone cloning the cpipe_util repository from github (you will need MGHA GitHub read access for this method):

git clone https://github.com/MelbourneGenomics/cpipe_util
cp cpipe_util/swift_credentials.sh cpipe

Or you can download the credentials file from NeCTAR (Compute → Access & Security → API Access → Download OpenStack RC File) and place it in the cpipe root directory. Then you'll need to rename the file as swift_credentials.sh. You will need MGHA NeCTAR access for this method.

Installation

Run ./install.sh and the full installation should take place.

If you have the swift credentials, this should take around an hour. If you do not, this could take more than a day, depending on the speed of your computer and its processing power

Install Script Options

The Cpipe installer has a number of command line flags that can be used to customise the installation. You can view the latest flags simply by running ./cpipe --help. Otherwise, some are documented here:

 --help, --usage
   Print this help page to stdout
 -n, --processes <process number>
   Set the maximum number of processes to use for the install. The higher number the faster the install, but the more memory used. Defaults to the output of 'nproc --all', the number of available processing units (currently 4 on your system)
 -c, --credentials </path/to/swift_credentials.sh>
   Use the specified swift credentials file to download assets from NeCTAR. Defaults to looking in the cpipe root directory
 -v, --verbose
   Print everything to stdout instead of just errors. Good for debugging an install
 -s, --no-swift
   Do a manual install instead of downloading assets from NeCTAR. Strongly NOT recommended as this will potentially take days to complete
 -t, --task <taskname>
   Specify one or more tasks to run instead of a full install. Each time you use this flag it will add another task. Don't use this unless you know what you're doing
 -p, --no-pip
   Don't update pip modules. Don't use this unless you know what you're doing

Set Installation Name (optional)

At this point you can set the name of your installation, which will become a prefix on all of the pipeline output files, ensuring they are unique to your installation.

To do this, create a file named pipeline_id in the root of your Cpipe installation, and inside the file write something like <sitename>_<version> where sitename is the name of your system/site, and version is the release version of Cpipe you are using. For example, vlsci_2.3. You can do this in with the following command:

echo "vlsci_2.3" > pipeline_id