Frequently asked questions (FAQ)¶

Last changed: 2024-04-17

Project quotas vs. flavors ¶

Quotas are operational limits. For example, the number of gigabytes allowed for each project can be controlled so that cloud resources are optimized. Using the Overview tab on Dashboard will show you the quotas for existing projects.

Flavors define the compute, memory, and storage capacity of computing instances. It is the size of a virtual machine/instance that can be launched.

Capacity planning and scaling ¶

Cloud-based applications typically request more discrete hardware (horizontal scaling) as opposed to traditional applications, which require larger hardware to scale (vertical scaling).

OpenStack is designed to be horizontally scalable. Rather than switching to larger servers, you procure more servers and simply install identically configured services.

Scalability: possibility to add more virtual machines/instances as needed.
Flexibility: easier to install, implement and debug.
Better performance: uptime and live migration.

HTTP 401 Unauthorized Error from the OpenStack API ¶

The request you have made requires authentication. (HTTP 401) (Request-ID: req-xxxx-xxxx-xxxx-xxxx-xxxx)

To get access to OpenStack services, you need to have an authentication token. A token represents the authenticated identity of your username, password, project, domain, etc.

Each API-request includes a spesific authentication token. To access multiple services, you need to have a valid token for each service. A token can become invalid for different reasons. E.g. if you have wrong username, password, domain, user role, or lacking proper access to a project. Administrative services such as openstack user, project, group, domain, etc. will also give you an unauthorized error.

Transferring a volume ¶

To transfer a volume from one project to another, both projects have to be within the same region. Please also note that the projects cannot use the same volume simultaneously.

You will experience Unable to accept volume transfer error if you try to transfer a volume to a project which is located in another region, or if the project recipient does not have enough quota to accept the volume request.

Default OpenStack features disabled in NREC ¶

Some features which are ON by default in generic OpenStack, are for various reasons disabled in NREC. The affected features are listed in the table below.

Disabled OpenStack Features¶
Feature	Description	Comment
Reboot	Reboot instance	Workaround: Shut down and start instance
Resize Instance	Make size of instance larger	This is disabled in Dashboard (Web GUI). Workaround: Use CLI
Suspend	Suspend Image and store state on disk	Workaround: Pause instance
Shelve	Stop instance and free resources whilst retaining state (incl. IP addresses)	Workaround: Create snapshot. IP addresses are not retained, though.

How to regenerate your public SSH key ¶

If your public SSH keys have been mistakenly deleted or disappeared from the dashboard, and you haven’t got local copies, it is trivial to regenerate and readd them.

Run the following command in your terminal:

ssh-keygen -y -f <path to your private key>

This will output the public key to stdout which may be stored in a new file or copied to the clipboard.

To readd a key, go to the NREC Dashboard and click on on Key Pairs -> Import Public Key

By using openstack rebuild function, you can start an instance from a new image while maintaining the same IP addresses, amongst other metadata.

$ openstack server rebuild --image <image> <server>

Efficiently creating filesystems on large volumes ¶

XFS/EXT4 formatting on a disk of large size (e.g. several TB) using mkfs will under normal circumstances take a long time. This is because mkfs discards (clears) all blocks in the format process. For normal disks, especially SSD drives, this is what you want. However, due to the nature of volumes in NREC discarding is not needed. In order to significantly speed up mkfs, run without discarding:

For XFS:

mkfs.xfs -K /dev/<device>

For EXT4:

mkfs.ext4 -E nodiscard /dev/<device>

The time difference is huge for large volumes. Without discarding, mkfs takes a few seconds compared to several minutes (or hours) with discarding turned on.

How to acknowledge the use of NREC ¶

If you have used our infrastructure services for computing or other needs, we appreciate if you include this in your acknowledgment.

An example of an acknowledgement of having used NREC is:

The computations were performed on the Norwegian Research and Education
Cloud (NREC), using resources provided by the University of
Bergen and the University of Oslo. http://www.nrec.no/

Transferring an instance between projects using snapshot ¶

While it isn’t possible to “move” an instance between different projects without interruption, you can utilize the snapshot feature to transfer an instance from one project to another. Note that resources such as security groups and volumes are not transferred with the snapshot, and must be reconstructed in the new project.

In order to transfer a workload between projectA and projectB, simply follow these steps:

Take a snapshot of the instance in projectA as descibed here: Creating a snapshot
Download the snapshot to a local computer: Downloading a snapshot
Upload the snapshot to projectB: Uploading a snapshot
Launch a new instance in projectB using your snapshot as the source image: Launching a snapshot
Optionally delete the instance and snapshot in projectB if they aren’t needed anymore.

These steps can be done without deleting the instance in projectA, i.e. you can verify that all is well in the new instance before deleting the old instance.

Note that you can’t reuse the IP addresses of the old instance when creating a new in projectB. The new instance will have a different set of IP addresses. Depending on the application, you may need to configure either the application itself, the clients, or both.

Tip

Since downloading a snapshot can only be done using the CLI, we recommend doing the whole operation in the CLI rather than using the dashboard.

Does NREC block any incoming network traffic?¶

Yes, certain ports are blocked completely or partially in order to protect our users and their services running on NREC. More details here:

ACL for Incoming Traffic