Explore possibility of a Fedora-less CLAW

[bryjbrown]

Here at FSU we are heavily invested in AWS. We have recently completed a project to create a test deployment of 7.x to AWS using CloudFormation so that all the running VMs are on separate EC2 instances and all storage is handled using AWS storage solutions. We've found that we really like this set up, and would like to stick with it for all future projects as its become a core part of our infrastructure and skill set.

This has led to us reassessing the benefits that Fedora 4 brings us when we are already relying heavily on AWS storage. We are still sold on Drupal 8 and Solr as core parts of a repository solution, but we are now realizing that many of the features Fedora offers aren't necessarily worth the complexity of running a Fedora for our use case, where a pure Drupal/Solr solution should be enough to achieve our goals without the complexity of the current CLAW architecture.

Since the current CLAW implementation seems to already be somewhat loosely coupled in terms of Drupal 8 and Fedora (my understanding is that its essentially a stand-alone Drupal 8 broadcasting messages through Camel that Fedora 4 uses the sync its state), and I have heard that there is now work to shift Solr to be directly connected to Drupal 8 instead of making it listen to the Camel queue, how hard would it be to ONLY run the Drupal 8 part, so that folks could use AWS storage on the Drupal 8 filesystem if they choose? Currently the only issue I can think of would be missing out on API-X, what else am I missing?

We think that there might be others in the community that feel the same way (sold on Drupal 8/Solr but not necessarily on the extra complexity of running Fedora), so I'm posting this issue as a place to discuss this question as well as a place for anyone who finds themselves in the same situation we are in to discuss their use cases.

[dannylamb]

@bryjbrown Solr currently is indexing off of Drupal, as well as the triplestore. And although you'd miss out on API-X, the microservices themselves work with just Drupal no problem. Generally speaking, it would look and run the same on the front-end without any less features and would still be scalable.

What you'd miss out on would be back-end features that are directly Fedora-related. The biggest one I can think of from a preservation perspective is the audit trail, so you'd definitely want to weigh that out. You'd also want to make regular backups/snapshots of your Drupal db, since you won't have another copy of the metadata laying around in Fedora.

It would be good to hear if there's others from the community considering similar setups, because this is an interesting use case. Although it's not the intention of the software to run without Fedora, from a technical perspective it would be easy to achieve because of CLAW's architecture.

[ajs6f]

This is of great interest to us, for a couple of reasons. One is that we have scaling needs that Fedora won't support (hence our interest in Trellis). That's still an LDP component, but a very different animal. The other is to be able to create very minimal deployments that can operate reliably in circumstances with very limited resources (e.g. a marine biology research ship out on the open ocean). That would be more like what @bryjbrown is talking about.

[mjordan]

Hello from the Boston subway, I have been working on a fixity audit microservice while traveling for the last two weeks. Needs cleanup and more tests but it will support querying a resource for fixity digests from a variety of sources, not just Fedora, so we can get it to work with standalone Drupal.

[ajs6f]

[ajs6f]

There's a potential useful (or maybe merely related) development discussed for Trellis here.

[ruebot]

1.5 or so years away from everything, I don't much see the use for fcrepo in the stack anymore. I'd be interested in exploring this more in my 21hrs a month a York :smile: