Spymemcache- Memcache/Membase Faileover

问题

Platform: 64 Bit windows OS, spymemcached-2.7.3.jar, J2EE

We want to use two memcache/membase servers for caching solution. We want to allocate 1 GB memory to each memcache/membase server so total we can cache 2 GB data. We are using spymemcached java client for setting and getting data from memcache. We are not using any replication between two membase servers.

We loading memcacheClient object at the time of our J2EE application startup.

    URI server1 = new URI("http://192.168.100.111:8091/pools");
    URI server2 = new URI("http://127.0.0.1:8091/pools");
    ArrayList<URI> serverList = new ArrayList<URI>();
    serverList.add(server1);
    serverList.add(server2);
    client = new MemcachedClient(serverList, "default", "");

After that we are using memcacheClient to get and set value in memcache/membase server.

Object obj = client.get("spoon");
client.set("spoon", 50, "Hello World!");

Looks like memcacheClient is setting and getting and value only from server1.

If we stop server1, it fails to get/set value. Should it not use server2 in case of server1 down? Please let me know if we are doing anything wrong here...

回答1:

aspymemcached java client dos not handle membase failover for particular node.

Ref : https://blog.serverdensity.com/handling-memcached-failover/ We need to handle it manually(by our code)

We can do this by using ConnectionObserver Here is my code :

public static void main(String a[]) throws InterruptedException{
            try {
                URI server1 = new URI("http://192.168.100.111:8091/pools");
                URI server2 = new URI("http://127.0.0.1:8091/pools");
                final  ArrayList<URI> serverList = new ArrayList<URI>();
                serverList.add(server1);
                serverList.add(server2);
               final MemcachedClient client = new MemcachedClient(serverList, "bucketName", "");
                client.addObserver(new ConnectionObserver() {

                    @Override
                    public void connectionLost(SocketAddress arg0) {
                        //method call when connection lost
                        for(MemcachedNode node : client.getNodeLocator().getAll()){
                            if(!node.isActive()){
                                client.shutdown();
                                //re init your client here, and after re-init it will connect to your secodry node
                                break;
                            }
                        }
                    }
                    @Override
                    public void connectionEstablished(SocketAddress arg0, int arg1) {
                        //method call when connection established 
                    }
                });
                Object obj = client.get("spoon");
                client.set("spoon", 50, "Hello World!");
            }  catch (Exception e) {
            }
    }

回答2:

client.get() would use first available node and therefore your value would be stored/updated on one node only.

You seems to be a bit contradicting in your requirements - first you're saying that 'we want to allocate 1 GB memory to each memcache/membase server so total we can cache 2 GB data' which implies distributed cache model (particular key is stored on one node in the cache farm) and then you expect to fetch it if that node is down, which obviously won't happen.

If you need your cache farm to survive node failure without losing data cached on that node you should use replication, which is available in MemBase but obviously you would pay the price of storing the same values multiple times so your desire of '1GB per server...total 2GB of cache' won't be possible.

来源：https://stackoverflow.com/questions/9003525/spymemcache-memcache-membase-faileover