Our actor has to keep an instance of Client. We don't use a connection pool, because we will use multiple actors for handing parallel requests to a database. Look at the following struct:

pub struct CacheActor {
    client: Client,
    expiration: usize,
}

The struct also contains an expiration field that holds the time-to-live (TTL) period. This defines how long Redis will hold the value.

Add a new method to the implementation that uses a provided address string to create a Client instance, and adds both the client and expiration values to the CacheActor struct, as follows:

impl CacheActor {
    pub fn new(addr: &str, expiration: usize) -> Self {
        let client = Client::open(addr).unwrap();
        Self { client, expiration }
    }
}

Also, we have to implement an Actor trait for SyncContext, just as we did when resizing the worker in Chapter 10, Background Tasks and Thread Pools in Microservices:

impl Actor for CacheActor {
    type Context = SyncContext<Self>;
}

Now, we can add support for messages to set and get cached values.