Integration of Curator/Zookeeper

[Oliver Komoll]

Hi Thomas,

i want to implement a service discovery for my microservices based on curator and zookeeper. I register my services on startup with there version and a list of the supported tenants (logical databases) at zookeeper. 

I do not want to implement the discovery on the client side, cause i need also a solution for the web clients. So i came to the idea to use membrane as a configurable API-Gateway.

Do you think it is possible to implement a replacment or enhancement for the clusters (Balancer) class to describe something like

<serviceProxy name="myService" port="8080">

    <path>/service</path>

    <balancer>

        <clusters>

            <cluster zkServer="localhost:2181" basePath="services"/>

        </clusters>

        <roundRobinStrategy/>

    </balancer>

</serviceProxy>




Also i need to integrate a version number and the tenant. The tenant is stored in a header field and the required version of the service must be configured in membrane - maybe with the help of a custom interceptor - so my path for zookeeper looks like /services/myService/V1/tenant1 What do you think -is this possible - is it better to embed and configure membrane in java and "inject" the discovered server list or to enhance membrane




TIA

Oliver

[Thomas Bayer]

Hi Oliver,
we are working on something very similar to your approach with the etcd registry. You can have a look at:

https://github.com/membrane/service-proxy/tree/master/core/src/main/java/com/predic8/membrane/core/cloud/etcd

The basic idea is that the endpoint address of a target comes from the registry ( zookeeper, etcd, ... ) and the Membrane router than uses that information to route to that endpoint.

We abstracted the publisher using spring, so you can just implement a different publisher.

Am 26.11.15 um 16:26 schrieb Oliver Komoll:

Hi Thomas,

I do not want to implement the discovery on the client side, cause i need also a solution for the web clients. So i came to the idea to use membrane as a configurable API-Gateway.

We came up with the same ;-)

Do you think it is possible to implement a replacment or enhancement for the clusters (Balancer) class to describe something like

<serviceProxy name="myService" port="8080">

    <path>/service</path>

    <balancer>

        <clusters>

            <cluster zkServer="localhost:2181" basePath="services"/>

        </clusters>

        <roundRobinStrategy/>

    </balancer>

</serviceProxy>

I would not use the loadbalancer to do that. I think the publisher is a nicer approach.

Also i need to integrate a version number and the tenant. The tenant is stored in a header field and the required version of the service must be configured in membrane - maybe with the help of a custom interceptor - so my path for zookeeper looks like /services/myService/V1/tenant1

I think you can combine the publisher with an interceptor.

What do you think -is this possible - is it better to embed and configure membrane in java and "inject" the discovered server list or to enhance membrane

You can embed it into your Java code or you can deploy it as a Web App or install it in a Docker container. It does not really matter for that problem.

Would be cool, if you can share your code so we get also support for zookeeper.

Cheers,
Thomas

TIA

Oliver

--
You received this message because you are subscribed to the Google Groups "membrane-monitor" group.
To unsubscribe from this group and stop receiving emails from it, send an email to membrane-monit...@googlegroups.com.
To post to this group, send email to membrane...@googlegroups.com.
Visit this group at http://groups.google.com/group/membrane-monitor.
For more options, visit https://groups.google.com/d/optout.

[Oliver Komoll]

Hi Thomas,

sounds great - do you have a configuration example ?
Couldn't find the interface for the publisher - found only  the implemented EtcdPublisher, so how can i inject my own registry implementation

TIA

Oliver

Am Donnerstag, 26. November 2015 16:53:36 UTC+1 schrieb Thomas Bayer:

Hi Oliver,
we are working on something very similar to your approach with the etcd registry. You can have a look at:

https://github.com/membrane/service-proxy/tree/master/core/src/main/java/com/predic8/membrane/core/cloud/etcd

The basic idea is that the endpoint address of a target comes from the registry ( zookeeper, etcd, ... ) and the Membrane router than uses that information to route to that endpoint.

We abstracted the publisher using spring, so you can just implement a different publisher.

Am 26.11.15 um 16:26 schrieb Oliver Komoll:

Hi Thomas,

I do not want to implement the discovery on the client side, cause i need also a solution for the web clients. So i came to the idea to use membrane as a configurable API-Gateway.

We came up with the same ;-)

Do you think it is possible to implement a replacment or enhancement for the clusters (Balancer) class to describe something like

<serviceProxy name="myService" port="8080">

    <path>/service</path>

    <balancer>

        <clusters>

            <cluster zkServer="localhost:2181" basePath="services"/>

        </clusters>

        <roundRobinStrategy/>

    </balancer>

</serviceProxy>

I would not use the loadbalancer to do that. I think the publisher is a nicer approach.

...

[Tobias Polley]

Hi Oliver,

the setup Thomas described is beginning to be documented.

With the lastest Membrane build from the github master branch (a simple "mvn package -DskipTests" should do the trick), examples/service-discovery-with-etcd contains the description of the complete example.

This consists of one example node using Membrane and etcdPublisher to publish node info into etcd, and another (possibly load balancing) Membrane using etcdBasedConfigurator to automatically setup serviceProxies for all node infos found in etcd.

Please note that you don't actually have to use etcdPublisher, if the node(s) can publish itself within etcd (step 1: by basically making a bunch of HTTP calls after startup, step 2: include a TTL and repeat those calls shortly before expiration).

Best, Tobias