Add redis cache

[finestructure]

Adds a redis service to app.yml.

[finestructure]

I'll ad hoc deploy this from the branch to see how it goes in dev.

[finestructure]

We'll need to clarify the situation around the ports before deploying this in prod. I don't think they're open due to the firewall sitting in front of it but it needs double checking.

We should be able to connect to the 8001 port for the insights GUI via an ssh tunnel/port forward. I did that i the past but not in a while - need to dig up the commands again.

[heckj]

If it’s a swarm, there may need to be firewalls opened between the hosts - I don’t think it’s available in prod based on the current network security groups.

[finestructure]

All the swarm hosts are in the same NSG and I'm pretty sure we only need to open 8001 (to Bastion) in case we want to use the Redis GUI.

I've verified that this is working as intended on dev (which is technically also a swarm albeit with just one node).

[finestructure]

Having some trouble getting this to work with service order. I suspect depends_on is looking for a health check and the redis container doesn't have one.

I think I know how to fix this but I'll leave it until tomorrow.

[heckj]

did some reading when I got settled, and depends_on without a health check only verified start order, as you presumed. I found this example of using a CLI heathcheck for redis:

redis:
image: redis:latest
ports: - "6379:6379"
healthcheck:
test: ["CMD", "redis-cli", "ping"]
interval: 10s
timeout: 5s
retries: 3
start_period: 10s

[finestructure]

Ok, this is now working:

[ INFO ] Redis set/get result: 2025-01-03 08:43:49 +0000 [component: server]

The issue was that docker swarm deploy is stuck on the v3 docker compose file format and didn't support the new keys I was trying.

And using just redis as a depends_on entry is problematic, because it seems that the default redis container doesn't implement a health check which this relies on. So what happens is that the server doesn't launch if that dependency is set because it doesn't see redis being healthy.

That'd be fixable but I've simply worked around it by putting the temporary test write in a Task { } with a delay.

We won't really have to worry about launch order for the real deployment so this is a non-issue.

[finestructure]

Ok, connecting to the redis GUI also works via

ssh -N -i .ssh/bastion_id_rsa -L 8001:d1:8001 docker-user@bastion

The port is only open to connections from bastion.

[finestructure]

The alerting: section was indented with 4 spaces rather than 2 spaces like the rest.