search

AmadeusITGroup / HttpSessionReplacer

Store JEE Servlet HttpSessions in Redis
MIT License
49 stars 33 forks source link

Handle CLUSTERDOWN message #15

Closed nbogojevic closed 6 years ago

nbogojevic commented 7 years ago

When CLUSTERDOWN message occurs, latest operation should retried instead of failing immediately. Best place to do this could be in src/main/java/com/amadeus/session/repository/redis/JedisClusterFacade.java.

See following explanation:

https://groups.google.com/forum/#!msg/redis-db/FDK59yE9pus/WgDQIfJujboJ

nbogojevic commented 7 years ago

Not sure this should be addressed in client. CLUSTERDOWN means that either cluster is fully down or one of the masters missing (partial coverage), other masters are aware of the issue and we might be waiting for failover.

We can disable this functionality of waiting for full coverage via redis.conf and following option:

cluster-require-full-coverage no
nbogojevic commented 7 years ago

Some of the issues during cluster failover have bee solved in #18

nbogojevic commented 6 years ago

Will keep behavior as it is.