Chapter 6. Deploying a single logical service across many sites for failover
A typical scenario for using Service Interconnect is to deploy a server process on two sites with the intention that if one site fails, the other site seamlessly processes any further requests. In this scenario the primary server responds to all requests while that server is available and traffic is only directed to the secondary server when the primary server is not available. The procedure describes two servers, however this technique works for many servers.
Prerequisites
- Two or more unlinked sites.
- A basic understanding of Service Interconnect and its networking model.
Procedure
-
Create sites by using
skupper init. - Deploy your servers on different sites.
Generate a token on the first site:
$ skupper token create token.yaml
This file contains a key and the location of the site that created it.
NoteAccess to this file provides access to the service network. Protect it appropriately.
Use the token on the cluster that you want to connect from:
To create a link to the first site:
$ skupper link create token.yaml --cost 99999
The high cost setting means that traffic is not directed to this site under normal circumstances. However, if there is no other server available, all traffic is directed to this site.
Expose the servers on the service network for both sites.
Create the service:
$ skupper service create <name> <port>
where
-
<name>is the name of the service you want to create. -
<port>is the port the service uses.
By default, this service is now visible on both sites, although there is no server available to process requests to this service.
NoteBy default, if you create a service on one site, it is available on all sites. However, if
enable-service-syncis set tofalseyou need to create the service on both sites.-
Bind the service with the server on both sites.
$ skupper service bind <service-name> <target-type> <target-name>
where
-
<service-name>is the name of the service on the service network -
<target-type>is the object you want to expose,deployment,statefulset,pods, orservice. -
<target-name>is the name of the cluster service
For example:
$ skupper service bind hello-world-backend deployment hello-world-backend
-
- You can use the console to check the traffic flow or monitor the services using your tooling. Clients can connect to either site, and the server on that site processes the requests until the server is not available. Further requests are processed by the server on the other site.
If the server on the original site becomes available, it processes all further requests. However existing TCP connections to the secondary or backup server will persist until those TCP connections are closed.
