Throw the problematic exception node when a request exception occurs

[wang-qijia]

In a cluster or shard mode, it is often difficult to determine which node caused the exception when a request exception occurs, making it unfriendly for troubleshooting. It would be ideal to obtain the exception node for error logging or metric purpose.

Should we wrap the exception node in JedisConnectionException in the Connection class in order to record the error log more friendly?

[sazzad16]

@wang-qijia Thank you for sharing your idea. It's a interesting one.

[yangbodong22011]

@wang-qijia Good suggestion, I prepare a PR #3470 to solve it.

[wang-qijia]

@yangbodong22011 Can we make the HostAndPort variables in the exception class so that the application layer can easily access them?

[yangbodong22011]

JedisClusterOperationException is used in scenarios where an exception occurs in the Cluster, but not all scenarios have a legal hostAndPort(e.g. processKey). I prefer to put it in the exception message when the hostAndPort exists.

[wang-qijia]

The Cluster API is just one of the options. There are other modes such as sharding and using the Jedis API. While the Cluster mode has some specific handling, it might be possible to handle other modes in the Connection class.

https://github.com/redis/jedis/blob/master/src/main/java/redis/clients/jedis/Connection.java#L165