...
Currently, only the Edmonton region (yeg) has two GPU-enabled nodes, each hosting 4 NVIDIA K80 graphics cards equipped with 2 GPU cores and 24GB of onboard RAM. In total, there are 16 GPU cores available. Because we have so few resources at the moment, we have created a Lease Process to allow users to launch instances as normal but with a twenty-four hour lease on the instance launched. At the end of the lease, the instance will be automatically snapshotted and terminated. While the automated shelving process takes a snapshot of your instance, these snapshots should be considered a safety mechanism and will be deleted periodically.
| Status | ||||
|---|---|---|---|---|
|
| Note |
|---|
If you would like to save your work for future use, we ask that users create their own |
...
snapshot and do not rely on the shelving process. You can use the snapshot to launch a new instance when the GPU resources are needed again. |
There is a possibility that when you try to launch a GPU-enabled instance, you will get an error indicating that there are “not enough resources”, which means all the GPUs are currently in use. At this time the Lease Process is on a first-come-first-served basis with no reservations, so if there are no GPU cores available users will need to try again later.
...
Log-in to the Rapid Access Cloud dashboard at https://cloud.cybera.ca.
In the left-hand panel under “Compute”, click “Instances”.
Click the name of a GPU-enabled instance you would like to extend the lease on.
| Status | ||||
|---|---|---|---|---|
|
...
Note the “Lease Date”; this will initially be set to the time and date the instance was created plus twenty-four hours.
Click the “Submit” button; the Lease Date will be set to twenty-four hours from the time you click Submit.
...
2. Execute the following command using the command line tools:
| Code Block |
|---|
nova unshelve <instance uuid> |
...