Changes

Local Note: the cluster will soon get large, fast global storage, at this point local persistent volumes will be phased out and probably not available anymore. Tensorboard monitoring should be used to import training data and store results done using service exports, as explained below, and log files not make use of your training. There are special local PVs for monitoring the training using Tensorboard. Host directories are meant for common training data sets stored permanently on the host. They are always read only.

Since anyone can mount global persistent volumes in the same namespace, they can and should be used to share datasets. The name of a PVC which contains a useful dataset should start with "dataset-" and be descriptive, so that it can easily be found by other users. Also, the root of the PVC should contain a README with informations about the dataset (at least the source and what exactly it is). Finally, it is probably good practice if other users of the dataset which are not the creator mount the volume readonly (by specifying "readOnly: true" after the mountPath in the pod's yaml).

 === Global dataset storage for large, static datasets === Every node has a link to a global repository "/raid/datasets" in its filesystem, which sits A note on very fast NVMe raid mounting. Currently (1.9 GB/s readwill change in the near future) and , ceph volumes can be either mounted read-ReadWrite by a single pod only in every container as in , or ReadOnly by multiple pods. Thus, the following example: <syntaxhighlight lang="yaml">todo</syntaxhighlight> Please see workflow for a static dataset is to create the page [[CCU:Global dataset storage|on global storage]] for PVC, then create a list of available datasets and pod to write all the method data to upload your ownit, then delete this pod and mount it read only from now on so it can be used in multiple pods.