feat(charts): add optional persistence storage to influxdb and grafana

[rimusz]

No description provided.

[deis-bot]

@lachie83, @bacongobbler and @vdice are potential reviewers of this pull request based on my analysis of git blame information. Thanks @rimusz!

[vdice]

@rimusz can you expand on the setup needed to run this? I attempted to test locally using a GKE-based k8s cluster (in combination with the monitor chart changes in deis/monitor#169):

$ helm upgrade smelly-badger workflow-pr/workflow --version v2.7.1-20170119174902-sha.287dd60 -f workflow/values.yaml
Error: UPGRADE FAILED: unable to decode "": [pos 181]: json: error decoding base64 binary '<base64-encoded JSON data>': illegal base64 data at input byte 0

(where I use the workflow chart artifact that uses the monitor PR chart as its sub-chart and I reference the local workflow/values.yaml to get the change in this PR)

As it seems I need to supply the key_json, I do so:

$ B64_KEY_JSON="$(cat /Users/vaughndice/.secrets/deis-sandbox-creds.json | base64 -w 0)"

$ helm upgrade smelly-badger workflow-pr/workflow --version v2.7.1-20170119174902-sha.287dd60 -f workflow/values.yaml --set gcs.key_json="${B64_KEY_JSON}",registry-token-refresher.gcr.key_json="${B64_KEY_JSON}"

...but then pods are stuck in CrashLoopBackOff. I suppose my missing ingredient is specifically enabling use of GCP PD for my cluster?

[rimusz]

ok, been testing myself on GKE and AWS (coreos) on k8s v1.4.x and v1.5.x and it looks like the StorageClass needs to be created before hand.
Not sure how to deal with that yet, as StorageClass is different for GCE/AWS, to make it part of Workflow charts, have in an example in docs?

[vdice]

@rimusz yes, I think a docs addition to this PR would be suitable in this case -- then, I can use those to guide testing.

[rimusz]

checked with @technosophos there is no way to check that some StorageClass exists or not yet.
which does not make an easy way to install StorageClass for now.
Ok, then I'm going to write a doc on persistence storage for influxdb/grafana and have there two examples of StorageClass for GCE and AWS.

[vdice]

Tested to success in combination with #713 and deis/monitor#169. (PV/PVCs successfully created; grafana dashboards and influxdb metrics persist through pod destruction, as expected.)