Tyk Gateway (docker) with multiple nodes - where does configuration data get stored?

[zeigernz]

Hi all,
I’m a new Tyk user and am very excited about the product!

I’m planning on using a clustered setup of a bunch of Tyk Gateway nodes (only gateway for now, no dashboard/analytics).

Ideally, I’d like to run a cluster of gateways as headless docker containers talking to a separate redis cluster.

What I can’t seem to figure out is:

• Where does the API definition, configuration data get stored when using a cluster of gateway docker containers? In Redis?
• I tried adding a bunch of api definitions and that worked great. Then, I got rid of that container and spun up another one that talks to the same redis store. This new container could not pick up any of the api definitions I added earlier. Does that mean that they were lost?
• If local files are used to save these configurations (rather than redis), what are your thoughts on using a shared non-ephemeral file system shared with all the gateway nodes so that the configurations can end up there and when new nodes come up, they already know what the configuration looks like.

Thank you.

[Martin]

@zeigernz

Where does the API definition, configuration data get stored when using a cluster of gateway docker containers? In Redis?

These are file-based in Community Edition, so you would need to mount them or use the Gateway REST APi to create them on each container deploy

@zeigernz

I tried adding a bunch of api definitions and that worked great. Then, I got rid of that container and spun up another one that talks to the same redis store. This new container could not pick up any of the api definitions I added earlier. Does that mean that they were lost?

See above, they are files we do not store anything but key data and analytics in redis

@zeigernz

If local files are used to save these configurations (rather than redis), what are your thoughts on using a shared non-ephemeral file system shared with all the gateway nodes so that the configurations can end up there and when new nodes come up, they already know what the configuration looks like.

You could do this - though it can be difficult to mount a shared volume efficiently in a container cluster, but it is totally possible.

An easier approach would be to re-create them using the rest API, or store them in S3 and then have them mounted on container start (maybe fork the container).

[zeigernz]

Thank you @Martin, appreciate your prompt response.