Refocusing CyVerse Atmosphere FAQ¶
CyVerse has refocused Atmosphere to better serve the growing need for container and cloud native development services. Read on for what this change means to you.
What has changed?¶
CyVerse has changed Atmosphere from a general purpose cloud computing environment to one that supports cloud-native development projects. In a cloud-native environment, you can more fully take advantage of the distributed, scalable, flexible nature of the public cloud, which means spending less time worrying about networks, servers, operating systems, etc. and more time on your research.
Why did Atmosphere change?¶
Because containers have become a powerful and preferred way for researchers to create, execute, and share analyses, largely replacing the capabilities of Atmosphere’s virtual machines, we are now directing resources towards the growing need for cloud-native services. In addition, the NSF recently extended CyVerse’s long-term partnership with Jetstream through 2025 to provide similar capabilities and interfaces with significantly larger CPU, GPU, and storage infrastructure, presenting an exciting option for our U.S. users.
What do I need to do?¶
If you are a US-based researcher using Atmosphere to develop VICE or other DE apps, you may continue to use Atmosphere for this purpose. Otherwise, users were asked to transfer their resources off Atmosphere volumes by the following dates:
|Dates||U.S. Users||International Users|
|July 1, 2020||NA||You will no longer be able to create an account.|
|August 1, 2020||NA||You will no longer be able to log in. All resources should be moved off Atmosphere by this date.|
|September 1, 2020||You will no longer be able to log in. All resources should be moved off Atmosphere by this date.||NA|
What if I did not move my resources off Atmosphere by September 1, 2020?¶
After January 1, 2021, CyVerse will transfer any unclaimed Atmosphere volumes to the Data Store, when possible. We will archive any remaining unclaimed images with data and delete associated volumes.
How do I get help moving my resources off Atmosphere?¶
CyVerse staff can help with planning your move, migrating your data and tools from Atmosphere to CyVerse’s other services, and advising you on alternative offerings for moving your data and tools; please contact us using the chat service in Atmosphere or email firstname.lastname@example.org.
What can I use instead of Atmosphere?¶
CyVerse staff will provide assistance for U.S.-based researchers to migrate their cloud images to Jetstream, which uses CyVerse Atmosphere as its primary interface. The CyVerse Atmosphere image must be owned by the user and must meet Jetstream’s requirements for importing images. CyVerse cannot guarantee 100% success when exporting a virtual disk image from CyVerse Atmosphere to Jetstream Atmosphere. Jetstream staff will not provide support for imported images, and instead recommends images be recreated in their cloud. For more information about Jetstream, see Getting Started with Jetstream.
What has happened to my community-created image?¶
Existing community-created images with an associated DOI archive still visible on the CyVerse Atmosphere image catalog after September 1, 2020, but may only be launched by special permission. Other community-created images may be archived and eventually will be removed from the image catalog.
Can I get a copy of my image?¶
Yes. To obtain a copy of the virtual disk of any image you have created/authored,please contact email@example.com by January 1, 2021. Although CyVerse cannot guarantee 100% success when importing a virtual disk image from Atmosphere, virtual disk images should be compatible with other OpenStack clouds.
What if I still want to use Atmosphere as a computing environment?¶
As stated above, Atmosphere has been refocused to support U.S.-based researchers’ cloud-native development projects.
Can I continue to use my Atmosphere images for a class or workshop?¶
You can use NSF’s Jetstream, which provides the same functionality as Atmosphere for launching images for classes and workshops. Contact firstname.lastname@example.org for assistance with this process if needed. Jetstream also will meet your needs if you want to use something more high-powered, such as Jupyter Hub, for your teaching sessions.
What can I use Atmosphere for after September 1, 2020?¶
Atmosphere will remain available as a support service for cloud native development projects, such as U.S. users who are developing containers. Contact email@example.com with a request.
Can I still use Atmosphere for single instances, such as launching a Jupyter Notebook?¶
CyVerse’s Visual Interactive Computing Environment (VICE) within the Discovery Environment will meet your needs for doing reproducible analyses using single instances.
Using Atmosphere FAQ¶
Where should I put save data on my Atmosphere Instance?¶
When doing work in Atmosphere, you will likely bring data into instances from the CyVerse Data Store, your local system, or another data source. However, within the instance, where should you put your data?
There are a several instance storage options within an instance, each with different considerations:
- Instance volume (usually mounted as
- Instance scratch directory,
- FUSE-mount your Data Store directory
- Home directory
Each solution will be briefly described below.
Larger storage, good performance, durable, not accessible outside of Atmosphere
Expand to read more
An instance volume, often referred to as simply “volume”, is a durable, long-term storage option that can only be used within Atmosphere instances. This is a reasonable storage solution for larger data sets that will be used within Atmosphere for many weeks or months. We recommend backing up your data to your Data Store home directory.
Note: A volume can be attached to one instance at a time, much like a virtual USB drive.
Instance Scratch Directory
Storage size according to instance size, high performance, ephemeral, not accessible outside of Atmosphere
Expand to read more
Every instance will have a ephemeral disk that is stored under
/scratch, and its storage size is defined by the size of the instance. The ephemeral disk only exists while the instance is active, and any data stored within this directory may be deleted when the instances is inactive (i.e. suspended, stopped, or shelved). This is a reasonable place for intermediate files or interim “work” before moving it to a long term storage (1. CyVerse Data Store, 2. an attached volume).
FUSE-mount your Data Store directory
Large storage, low performance, extremely durable, accessible with CyVerse’s ecosystem
Expand to read more
You can mount your Data Store home directory within your instance through FUSE. This method of access is convenient within certain limits. Performance may degrade if there are too many concurrent file access (e.g. several dozen concurrent files accessed simultaneously) or the files are too large (e.g. 100 Gb+).
Small storage, high performance, ephemeral, and not accessible outside of Atmosphere
Expand to read more
/homedirectory, the directory which you see when you login to an instance, will have limited disk space, whatever is configured for the image. This directory should also be considered temporary, if the instance is deleted or imaged. Unlike the scratch directory, the home directory will persist if the instance becomes inactive.
Viewing attached volumes and scratch
A command that will help see what is mounted & available is
df. The example
below uses the
-h option for storage reporting in more human readable form,
see details Below
Expand to read more
$ df -h Filesystem Size Used Avail Use% Mounted on /dev/vda1 20G 2.6G 17G 14% / none 4.0K 0 4.0K 0% /sys/fs/cgroup udev 7.9G 12K 7.9G 1% /dev tmpfs 1.6G 412K 1.6G 1% /run none 5.0M 0 5.0M 0% /run/lock none 7.9G 0 7.9G 0% /run/shm none 100M 0 100M 0% /run/user /dev/vdb 158G 60M 150G 1% /scratch /dev/vdc 64G 52M 61G 1% /vol_c
This instance has a 65 GB attached volume mounted at
/vol_c and over 150GB
One approach to deciding might be:
/scratch” would fit if my analysis likely to “start” with the large
dataset (30 fastq of 150 GB). Then, various tools are run that produce new data.
Once those are complete, only a smaller set of outputs will be retained. In this
case, we use
/scratch for computing intermediate files and only keep outputs.
“Attaching a volume” would fit if my analysis likely to begin with a common input. Then, an analysis (or many) are performed. Once complete, the results are the input to “periodic” analysis or refined over time.
In the end, any of the storage options mentioned in this article can be used for performing analysis. It is a matter of evaluating what data needs to be retained or the performance requirements. The CyVerse Data Store, or a Volume, are the most durable options for keeping data.
How do I manage disk space on my Atmosphere Instance?¶
A recent change in Atmosphere “relocated” ephemeral storage mounted within the instance’s filesystem.
Previously, Atmosphere would mount the ephemeral storage under the home
/home. This allowed you to automatically have a larger temporary,
working space when you established a remote connection via SSH or VNC.
This temporary space is relative to the “instance size” you selected on launch.
Due to the nature of ephemeral storage, namely that this storage may be erased
when the instance is not active, Atmosphere now mounts this storage onto
/scratch. You are welcome to use /scratch if you need additional storage for
active analysis or work.
See details below:
Expand to read more
$ df -h Filesystem Size Used Avail Use% Mounted on udev 16G 0 16G 0% /dev tmpfs 3.2G 8.7M 3.2G 1% /run /dev/vda1 20G 4.9G 15G 26% / tmpfs 16G 0 16G 0% /dev/shm tmpfs 5.0M 0 5.0M 0% /run/lock tmpfs 16G 0 16G 0% /sys/fs/cgroup /dev/vdb 237G 15G 210G 7% /scratch tmpfs 3.2G 0 3.2G 0% /run/user/0
It is important to note is that data under
/scratch will not persist when
shelving an instance. Also, instances which are in a suspended or stopped
state may ultimately be shelved if the instance is inactive for too long.
This consideration should be weighed when choosing where to put data. If a
higher durability guarantee is required (or desired) within the instance,
then store data in a volume. You can also backup and restore your data to
the CyVerse Data Store.