Replication should only execute deterministic commands

[adiholden]

f.e
spop

[romange]

Also RENAME should be implemented as "RESTORE" and "DEL", imho

[dranikpg]

Why should you encode the whole value just to rename it? Doesn't that slow down the whole flow a lot?

[romange]

Because how we allocate from heap per thread. If we move objects without copying it messes up stats. Rename is a management command and it can be slow

…

On Fri, Jan 6, 2023, 08:41 Vladislav ***@***.***> wrote: Why should you encode the whole value just to rename it? Doesn't that slow down the whole flow a lot? — Reply to this email directly, view it on GitHub <#604 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA4BFCAMCFGIH33OGDCUQKDWQ65DFANCNFSM6AAAAAATJRELFI> . You are receiving this because you commented.Message ID: ***@***.***>

[adiholden]

More commands to rewrite
SDIFFSTORE
SINTERSTORE
SMOVE
SPOP
SUNIONSTORE
BITOP
RENAME
RENAMENX
BLPOP
BRPOP
LMOVE
MOVE - current impl runs as global and written to all shards

[adiholden]

The non deterministic commands that need to be rewrite:
spop
blpop
brpop

[dranikpg]

You already listed them 😅 Lets exclude the blocking ones in this issue and strike through the ones that were implemented: spop and {p}expire{at}

[adiholden]

@dranikpg I think you finished this one right?

[dranikpg]

I think yes