Clear Redis Cache after reingest

[vgeorge]

Currently, the Redis cache is set to live for week. If a model update or scenario reingest is made, there will be no changes on the Explorer until the cache is expired. I can think of two approaches to fix this:

1. Implement a admin route to flush all Redis cache;
2. Add ingestedAt to the model record and do not return cached values if ingestedAt > timestamp of query cache.

Any thoughs, @danielfdsilva?

[danielfdsilva]

It seems that Redis does not store the cache timestamp creation, so we'd need to add that to our stored data if we are to implement 2.

1 is nice since it allow us to flush the cache at will, and we can do this by model, so we don't have to delete it all.

Another approach would be to connect to redis and clear caches for the models being imported. This one would require exporting another connection string when using the CLI.

[vgeorge]

I would prefer the approach 2 because it is seamless to the user and guarantees that the cache is in sync with the database. In option 1, the maintainer has to remember to flush cache. Option 3 adds complexity to the setup and is prone to connection errors to the Redis instance.

[vgeorge]

While this isn't fixed yet, I'm using AWS CLI to reboot the instances with by running following locally:

aws elasticache --region us-west-2 reboot-cache-cluster --cache-cluster-id gep-001 --cache-node-ids-to-reboot 0001
aws elasticache --region us-west-2 reboot-cache-cluster --cache-cluster-id gep-002 --cache-node-ids-to-reboot 0001
aws elasticache --region us-west-2 reboot-cache-cluster --cache-cluster-id gep-003 --cache-node-ids-to-reboot 0001

cc @EricSoroos