- reduce node connection default memory allocation to 512k.
- fix memcache node connection bugs
- reduce prometheus memory usage when handles connection errors.
- set network io err to unify error in case of mem boom
- add reload feature for cache type: redis, memcache, memcache_binary
- add more metrics
- add fuzz test for redis/memcache protocol
- add migrate tools named anzi
- add cluster manager tools enri
- fixed overlord panic when all cluster seed nodes down
- add back prometheus monitor metrics
- change pinger as long connection.
- ping timeout 100ms.
- fix max connection limit.
- fix pipe read when one err.
- max redirects 5.
- reset sub message only in need.
- refactor message pipeline.
- non-persistent connection for cluster redirect.
- add redis cluster support.
- change round-chan to race-chan.
- hot fix reconn.
- add executor interface.
- add pinger interface.
- add MsgBatchAlloctor for one node WR.
- fix batchdone err.
- add node reconn.
- fix message and request reuse bug.
- add redis protocol support.
- add memcache binary protocol support.
- add conf file check.
- compitable consist hash with twemproxy.
- reduce object alloc (by using Pool).
- recycle using of request and response memory.
- batch message execute to reduce conflict of command channel.
- using writev to reduce syscall times.
- synchronously send and wait of message.
- zero copy of request and response data at all.