RPUSH commands infinitely repeated with follower-proxy

[aphyr]

With Redis f88f866 and Redisraft 1b3fbf6, executing a single RPUSH command against a fresh cluster creates an infinite loop, wherein that RPUSH command is executed repeatedly, gradually ballooning the size of the value, state machine, and log. It also seems to cause some (all?) redis-cli requests against the node which executed the RPUSH to hang indefinitely.

For example, try this command with a fresh cluster:

/opt/redis/redis-cli rpush "a-list" "x"
(integer) 1

redis-cli monitor shows the Raft log applying endless RPUSH operations:

1583259973.634023 [0 ?:0] "RAFT" "rpush" "a-list" "x"
1583259973.634067 [0 ?:0] "RAFT" "rpush" "a-list" "x"
1583259973.634087 [0 ?:0] "RAFT" "rpush" "a-list" "x"
1583259973.634111 [0 ?:0] "RAFT" "rpush" "a-list" "x"
1583259973.634153 [0 ?:0] "RAFT" "rpush" "a-list" "x"
1583259973.634181 [0 ?:0] "RAFT" "rpush" "a-list" "x"
1583259973.634208 [0 ?:0] "RAFT" "rpush" "a-list" "x"

In this Jepsen run, we execute multiple RPUSH operations, each with a distinct integer value: you can see their consequences interleaved in reads:

[:r 0 [1 1 1 1 1 2 3 1 1 1 1 1 1 1 4 1 1 1 1 1 1 1 1 1 5 9 10 6 2 2 2 3 3 2 3 3 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 4 4 4 1 1 4 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 7 1 1 1 1 1 1 1 1 1 1 1 8]

This bug appears linked to follower-proxy=yes; running without follower proxies seems to resolve the issue.

[yossigo]

This should be fixed in #15.

[aphyr]

Yup! Looks good as of d589127