Huge lists do not migrate fully between cluster nodes

[chakaz]

(this may apply to other data types as well)

To reproduce:

• Run 2 cluster-mode Dragonfly processes (with with --port=7001 and --port=7002, and pass --cluster_mode=yes to both)
• Configure them to have all slots belong to the first:

./cluster_mgr.py --action=config_single_remote --target_port=7001
./cluster_mgr.py --action=attach --target_port=7001 --attach_port=7002

• Debug populate a huge list on the node with the slots:

$ redis-cli -p 7001
localhost:7001> debug populate 1 l: 1000 RAND TYPE list ELEMENTS 500000
OK
localhost:7001> llen l::0
(integer) 500000

• Migrate all slots to second node:

./cluster_mgr.py --action=migrate --slot_start=0 --slot_end=16383 --target_port=7002

• Note the issue:

localhost:7002> llen l::0
(integer) 4096
localhost:7002> memory usage l::0
(integer) 86064

[anadion]

We hame similar problem with RENAME command and big zset keys.

redis_version:6.2.11
dragonfly_version:df-v1.24.0
redis_mode:standalone

127.0.0.1:6379[14]> type ipv4
zset

127.0.0.1:6379[14]> ZCARD ipv6_tmp
(integer) 7391471
127.0.0.1:6379[14]> RENAME ipv6_tmp ipv6
OK
(6.75s)
127.0.0.1:6379[14]> ZCARD ipv6
(integer) 4092

[chakaz]

Yes @anadion, that is due to the same root cause. Thanks for reporting!