User shall have ROLE_ADMIN to configure system-level settings.
In certain cases jobs may fail with unexpected errors if the compute node runs Out of memory.
Admin users can configure a default swap file to the compute node being created. This allow to avoid runs failures due to memory limits.
To configure the size and the location of the swap:
- Open the Settings pop-up
- Click the Preference tab
- Select the Cluster section
- Click in the field under the
cluster.networks.config
label
- Insert the similar
json
block into a region/cloud specific configuration:
Where:swap_ratio
- defines a swap file size. It is equal the node RAM multiplied by that ratio. If ratio is 0, a swap file will not be created (default value: 0)swap_location
- defines a location of the swap file. If that option is not set - default location will be used (default: AWS will useSSD/gp2
EBS, Azure will use Temporary Storage)
- Click the Save button
- Click the OK button
- Now, while launch any pipeline, you can see specified swap settings in the run logs:
- To check that settings were applied, open SSH session and input the
swapon
command: