| fair-software.nl recommendations | Badge |
|---|---|
| 1. Code Repository |  |
| 2. License |  |
| 3. Community Registry |  |
| 4. Enable Citation |  |
| 5. Code Quality Checklist |
Automated instantiation and deployment of (clusters of) virtual machine(s) on bare metal using the OpenNebula platform, as well as subsequent provisioning and deployment of services incl., e.g. Dask.
marzipan consists of the core marzipan.py python module providing a high level interface to the OpenNebula cloud,
as well as an accompanying Docker framework and configurable deployment scripts providing a fully automated instantiation and provisioning environment.
For provisioning marzipan makes use of the emma_marzipan fork ansible playbooks.
marzipan is based off and strongly draws from Lokum, but is updated to make use of current versions of Ansible as well as python 3, and circumvents recurrent synchronicity and timeout issues arsing from the interplay of terraform, the runtastic OpenNebula provider for terraform, and various (legacy) OpenNebula versions.
marzipan has been tested on the SURFsara HPC cloud, but should work for any OpenNebula platform.
To make use of marzipan the user should clone this repository to their local system. Further instructions on the use of marzipan assume a full replica of the repository on the users local system.
The user should modify the ClusterConf.ini file located in the config subdirectory, as well as the opennebula_goera.tpl file in the templates subdirectory to match their requirements.
The ClusterConf.ini file enables the user to set desired configuration values such as the number of nodes, the name of the VMs, the OpenNebula endpoint and their credentials.
The config subdirectory of the repository includes a file ClusterConf.ini.example which can be appropriately modified and subsequently renamed.
The user must supply a template file specifiying the desired configuration for the VM(s) to be created.
An example, opennebula_goera.tpl, is provided in the templates subfolder of the repository.
In particular, the following fields will require modification:
CONTEXT = [
GROUP = "your_group"]
DISK = [
DATASTORE = "nameOfYourBaseImageDataStore"
DATASTORE_ID = "IDOfYourBaseImageDataStore"
IMAGE_ID ="IDOfYourBaseImage"
]
Please bear in mind, that the base image for the OS disk must be made available for the user (with the credentials being used) before executing marzipan. This is up to the user and can be accomplished using the OpenNebula user interface.
Change directories to the Docker subdirectory.
Build the nlesc/marzipan docker image by running
./build_marzipan.shThis creates the image with tag set to latest.
Change back to the root directory of the repository. The docker framework can then be used to instantiate and provision a cluster of VMs by running
./deployCluster.sh. The root and ubuntu user ssh keys generated for the cluster, as well as the hosts.yaml file enabling provisioning with the emma platform leveraging ansible are written to the deployments subdirectory (is created on execution) in a subfolder with the clusters name. They can be used to subsequently interact with the cluster.
The user can adapt the provisioning by modifying the marzipan_deploy.py script in the marzipan_scripts subdirectory in the section below
"""
emma based provisioninig
"""
The user is referred to the emma_marzipan fork for supported options.
NOTE: changes to the marzipan_deploy.py script require the docker image to be rebuilt before taking effect.
The cluster VMs can be accesed via ssh as ubuntu or root user, using the generated keys. For example:
ssh -i ./deployments/<clustername>/id_rsa_marzipan_root.key root@SERVER_IP
or
ssh -i ./deployments/<clustername>/id_rsa_marzipan_ubuntu.key ubuntu@SERVER_IPThe core marzipan.py module (located in the Marzipan subfolder) provides the one_interface class. The class' methods provide a high level interface to set up a cluster of VMs on the (SURFsara) OpenNebula cloud.
marzipan.py can be imported, providing access to the class methods, or run as a script to fully automatedly set up a cluster of VMs. marzipan.py also provides a high level class method deploy_cluster() which corresponds to the execution as a script.
marzipan is complemented by a ClusterConf.ini file (in config), where the user can set desired configuration values such as the number of nodes, the name of the VMs, the OpenNebula endpoint and their credentials. The repository includes a file ClusterConf.ini.example which can be appropriately modified and subsequently renamed.
Furthermore, the user should supply a template file specifying the desired VM configuration. An example, opennebula_goera.tpl, is provided in the templates subfolder of the repository. Please bear in mind, that the base image for the OS disk must be made available for the user (with the credentials being used) before executing marzipan.
Finally, the user can provide a public ssh key file for the root user to be included with the template. Alternatively, the ssh key can be supplied in the ClusterConf.ini file.
NOTE: if using the Docker framework, the use should refrain from, or take great care in, changing the settings relating to the ssh keys, as these are autogenerated during deployment.
When run as a script or by invoking the full deployment method mazipan will construct a VM template in OpenNebula, and deploy the requested number of VMs based on this template. marzipan will then monitor the deployment, only reporting successful execution when all VMs are in the RUNNING LCM_STATE. If this has failed to complete after 120 seconds marzipan will exit, notifying the user of failure.