Add external tool for Open OnDemand to Harvard Dataverse and add banner

[pdurbin]

We want to add a dataset-level tool that says "Explore in OpenOnDemand" to Harvard Dataverse, similar to how we added it to demo:

• To demo server, add external tool for Open OnDemand #390
• add Open OnDemand external tool manifest for demo #391

I believe this is the manifest we should use for Harvard Dataverse (the only difference from demo is the toolUrl, which is for a prod-specific landing page):

curl --location 'http://localhost:8080/api/admin/externalTools' \
--header 'Content-Type: application/json' \
--data '{
  "displayName": "Explore in Open OnDemand",
  "description": "Download dataset files into Open OnDemand and publish back to Dataverse",
  "toolName": "ondemand_loop_dataset_tool",
  "scope": "dataset",
  "types": ["explore"],
  "toolUrl": "https://hmdc.github.io/fasrc-ondemand-loop/dv_external_tools/landing.html",
  "httpMethod": "GET",
  "toolParameters": {
    "queryParameters": [
      {"dataverse_url": "{siteUrl}"},
      {"dataset_id": "{datasetPid}"},
      {"version": "{datasetVersion}"},
      {"locale": "{localeCode}"},
      {"env_cannon_prod": "{localeCode}"}
    ]
  }
}'

Once we've added it to Harvard Dataverse, let's capture it in a PR similar to #391 but this time for prod instead of demo.

For the banner, @sbarbosadataverse and @cmbz approved the following text:

"Integration with Harvard's Open OnDemand installation has been enabled. This allows Harvard affiliates with a FASRC account who are on the FASRC VPN use this HPC (high performance computing) integration."

[cmbz]

@landreev and @stevenwinship just a head's up: this work will need to be performed on Wednesday, 17 September. Can one of you plan to be available that day to perform these steps?

[pdurbin]

In theory, problems with the Loop deployment at FASRC could cause a delay, but we aren't expecting any.

[landreev]

@pdurbin Feel free to just add the tool yourself, when everything's ready
I haven't gotten around to playing with it on demo. Could you walk me through what exactly this thing does in terms of data access? Does the user select the specific datafiles they want to download; or will it just try to download the entire dataset at once?

[pdurbin]

@landreev users select specify datafiles and they are downloaded one by one. Don't worry, no on-the-fly zipping on the Dataverse server. Because we have S3 direct download enabled for Harvard Dataverse, it'll be used.

[landreev]

Thanks. Cavalier on-the-fly zipping of entire datasets was one of potential concerns. 😄 The other possible issue is that a large number of downloads in quick succession (for a dataset with a large number of smaller files) would get slapped with HUIT rate limit restrictions. If this becomes an issue, it should be solvable, since all these requests will be coming from FASRC ips that HUIT could (hopefully) whitelist. If worse comes to worst, I'm assuming it should be possible to add some self-throttling to the code that runs these downloads (similarly to what Jim has added to his download and upload tools).

[landreev]

Once again, this is trivial, we can activate the tool whenever.
But, if you need it ASAP and neither I nor @stevenwinship are around, feel free to

• log in on dvn-cloud-app-2;
• run /usr/local/dataverse-admin/support/ondemand/enable.sh
  this will add both the external tool and and the banner