Reading and Writing Data on a Persistent Volume
The two preceding examples staged their data with ingress and cleared it away when the
ephemeral volume was
destroyed. A persistent volume can remove the need for data
ingress and
egress altogether: the data a
workflow needs may already be
there when Fuzzball mounts it, and
anything a job writes survives the workflow. In the following example, we will mount a
persistent volume in the
job named new-file. The
volume already has the GitHub repo Hello-World
saved on it. The job copies the README file to a new file and appends text.
This example assumes that an administrator has already configured a storage provisioner with a persistent volume available to your group.
version: v4
volumes:
per_vol_1:
use: my-provisioner
name: my-persistent-volume
jobs:
new-file:
image:
uri: docker://alpine
mounts:
/data: per_vol_1
cwd: /data/Hello-World-master
script: |
#!/bin/sh
cp README README.new && echo "(goodbye dear basement)" >>README.new
resource:
cpu:
cores: 2
affinity: NUMA
memory:
size: 2GB
The example above uses the V4 syntax, where
usenames the provisioner andnamenames the volume on it. The previous (V1) syntax is still supported for backward compatibility. In V1, the equivalent would have been:volumes: per_vol_1: reference: volume://user/<storage-class>/<volume-name>Fuzzball rewrites that to
use: <storage-class>andname: <volume-name>when the workflow is submitted, and warns if no provisioner of that name exists. The third path segment is what makes it persistent – a two-segmentvolume://user/<storage-class>reference upgrades to an ephemeral volume on that provisioner.See Storage Volumes and Workflows for the full V4 volume syntax.
After the workflow completes successfully, you will find a new file on the persistent volume at
/Hello-World-master/README.new with the following contents:
Hello World!
(goodbye dear basement)