redis - Redis 集群重新分片后，许多不连续的槽分配给一个节点

Question

我创建了一个包含三个节点的 redis 集群:

~ redis-3.2.10/src/redis-trib.rb create 127.0.0.1:7000 127.0.0.1:7001 127.0.0.1:7002

然后我添加了第四个节点并重新分片以使时隙分配均匀:

~ redis-3.2.10/src/redis-trib.rb add-node 127.0.0.1:7003 127.0.0.1:7000
~ redis-3.2.10/src/redis-trib.rb reshard --timeout 60000  127.0.0.1:7000
...
How many slots do you want to move (from 1 to 16384)? 4096
...
Source node #1:all
...
redis-3.2.10/src/redis-trib.rb check 127.0.0.1:7000
>>> Performing Cluster Check (using node 127.0.0.1:7000)
M: 29fbb7e5f93eac22a224c14d4070139919bb0a5e 127.0.0.1:7000
   slots:1365-5460 (4096 slots) master
   0 additional replica(s)
M: c90c9f590bb48d328f3eed2fc96af3a7d9cb0f25 127.0.0.1:7003
   slots:0-1364,5461-6826,10923-12287 (4096 slots) master
   0 additional replica(s)
M: 970b3145574a4f38bbe10548bf6b80f8fdc2854d 127.0.0.1:7001
   slots:6827-10922 (4096 slots) master
   0 additional replica(s)
M: 2cc0922fcfb3e619b6d733e054dd249a4b6137bf 127.0.0.1:7002
   slots:12288-16383 (4096 slots) master
   0 additional replica(s)
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.

然后，如果您缩减规模，将插槽从 :7003 节点重新分片到其余三个节点，您最终会得到如下结果:

>>> Performing Cluster Check (using node 127.0.0.1:7000)
M: 29fbb7e5f93eac22a224c14d4070139919bb0a5e 127.0.0.1:7000
   slots:1365-5460,12288-13652 (5461 slots) master
   0 additional replica(s)
M: 970b3145574a4f38bbe10548bf6b80f8fdc2854d 127.0.0.1:7001
   slots:0-1364,5461-6826,10923-12287,13653-15017 (5461 slots) master
   0 additional replica(s)
M: 2cc0922fcfb3e619b6d733e054dd249a4b6137bf 127.0.0.1:7002
   slots:6827-10922,15018-16383 (5462 slots) master
   0 additional replica(s)
M: c90c9f590bb48d328f3eed2fc96af3a7d9cb0f25 127.0.0.1:7003
   slots: (0 slots) master
   0 additional replica(s)
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.

:7001 节点分配了 4 组不同的哈希槽。在随后的重新分片之后，哈希槽的这种“碎片化”似乎增加了。

哈希槽的碎片越来越多会不会成为一个问题(对性能或其他方面)？

Answer 1

AFAIK 否 - 哈希槽的“碎片”在任何有意义的开销方面都可以忽略不计。一些直觉:

• 小范围 (16K)
• 重新分片不是一个频繁的操作
• 在内部，Redis 使用位图来管理槽映射，因此无论实际槽的分布如何，访问这些位图都是在恒定时间内完成的。