双Tomcat集群共享缓存与故障互备方案咨询

Hey there, let's work through this cache sharing and failover problem you're facing with your two Tomcat instances behind Apache. I’ve helped troubleshoot similar setups before, so here are some actionable solutions to get you sorted:

The most reliable fix is moving your cache out of the Tomcat instances themselves and into a standalone, shared distributed cache. This way both Tomcats pull from the same data pool, regardless of which node handles the initial request.

• Redis: This is the go-to choice for most teams. It supports persistence, master-slave replication, and high availability. You can replace your local cache (like Guava Cache or Caffeine) with Redis using libraries like Jedis or Spring Data Redis. Even if one Tomcat goes down, the other will still pull the cached data directly from Redis.
• Memcached: A lightweight, in-memory distributed cache that’s great for high-throughput scenarios where persistence isn’t a requirement. It’s easy to integrate and ensures both Tomcats access the same cache dataset.

If you prefer to stick closer to Tomcat’s native tools, these options work well for two-node setups:

• Shared File System: Mount a networked storage volume (like NFS or SAN) to both Tomcat servers, and configure your cache to read/write from this shared directory. Just be sure to handle file locking properly to avoid concurrent write conflicts.
• Tomcat Cluster Cache Replication: Enable Tomcat’s built-in cluster functionality to replicate cache data between the two nodes. When one Tomcat writes to its cache, the data gets synced to the other instance. Note that this can have minor performance overhead, but it’s manageable for a two-node cluster.

If you don’t want to add a new service to your stack, repurpose your existing database as a shared cache layer:

• Create a dedicated table for cached data, with columns for the cache key, value, and expiration timestamp. Both Tomcats read from and write to this table. While it’s not as fast as a dedicated cache service, it’s a low-effort solution for smaller cache datasets.
• Don’t forget to add indexes to the cache key column to speed up lookups, and use database transactions to maintain data consistency.

Even with shared cache, you need Apache to reliably route traffic to the healthy Tomcat node:

• Configure mod_proxy and mod_proxy_balancer with health checks to detect failed nodes automatically. Here’s a quick example snippet:

<Proxy "balancer://tomcatcluster">
    BalancerMember http://tomcat1:8080 route=tomcat1
    BalancerMember http://tomcat2:8081 route=tomcat2
    ProxySet lbmethod=byrequests
    ProxySet ping=10
    ProxySet failonstatus=500,502,503,504
</Proxy>

This setup pings each node every 10 seconds and removes any node that returns critical error codes, ensuring all requests go to the healthy instance until the failed node recovers.

Quick Notes to Keep in Mind #

• Cache Consistency: No matter which solution you pick, make sure updates to cached data propagate correctly to avoid "stale" data being served. For distributed caches, consider using cache invalidation patterns instead of direct updates where possible.
• High Availability for Cache: If you go with a distributed cache like Redis, set up a master-slave or cluster configuration to avoid the cache itself becoming a single point of failure.

内容的提问来源于stack exchange，提问作者FDFDGSFG DJDBM