hydra-router require guarantee of running node during route match

[jkyberneees]

The router seems to be accepting the first URL match without guarantee that such service have a running node instance. In my case I just started a hapi-service-test service and the response from the router is this:

{
 "statusCode": 503,
 "statusMessage": "Service Unavailable",
 "statusDescription": "The server is currently unable to handle the request due to a temporary overloading or maintenance of the server. The implication is that this is a temporary condition which will be alleviated after some delay",
 "result": {
   "reason": "HR: [ol0r2s8oml] Unavailable express-service-test instances"
 },
 "tracer": "ol0r2s8oml"
}

Meaning that the hydra-router is considering old routes from the express-service-test service that I have ran in the past.

Regards, Rolando

[cjus]

@jkyberneees on startup, your hapi-service-test should have sent a message to hydra-router with its updated routes.

[jkyberneees]

Hi @cjus, it does it. However the issue appears when multiple services have registered the same route.
IMO the router should decide for the service that match the route but also is online.

[cjus]

@jkyberneees I looked into this. I don't see a way to address it. I think it's up to the service builder to ensure that routes are not reused by other services. Meaning two different services should not both register v1/{serviceName}/endpoint. I also think that service registries have to expire at some point. Perhaps a router that ends up with a 503 might flush the service registry or perhaps a Lua script in Redis.

[jkyberneees]

Hi @cjus, I can suggest you to check for service presence before accept the route match.
You can see an example here: https://github.com/jkyberneees/hydra-plugin-http/blob/master/libs/proxy.js#L35
I agree on the flush of the routes if 503 is detected, that would keep the router more performant and consistent.

Regards.

[arn-the-long-beard]

Hello
This discussion is old, but I have the same issue with hydra-router.

I used the different command for clean and refresh:
hydra-cli refresh
hydra-cli refresh node list

but hydra-cli routes show me the old dead services and then hydra-router cannot work correctly.

Is it something I am missing for cleaning/refresh ?

[cjus]

@arn-the-long-beard which version of hydra-cli are you using? How comfortable are you with Redis? Is this happening locally or on a production server?

[arn-the-long-beard]

Hello @cjus :)

I am using hydra-cli@1.5.5

For now this is happening locally. I am a very newbie to Hydra and Redis also. I am building micro services for a distributed system for the startup I am working.

[emirhg]

Same problem here. I'm using hydra-cli@1.6.0

hydra-cli refresh and hydra-cli refresh node listkeeps old routes, is not cleaning properly