If you have a slave server with a maxmemory value lower than its master, that slave will silently drop all writes that would exceed its maxmemory value. Even at the highest log levels, Redis does not output anything to indicate that it is ignoring writes. This can lead to significant consistency issues on the slave.
I'm not sure what the best solution is here, but I think at the very least, a slave should refuse to attach to a master if its maxmemory value is not the same as the master.
(This is reproducible on 2.6.16)
Comment From: charsyam
@tysontate hi. even though they have same strategy. it can cause same trouble. because load is not the same between master and slave.
I suggested this. master use maxmemory policy. and slave doesn't use maxmemory policy. at that time. some expiration will affect to slave. so It will keep consistency. good luck.
Comment From: enjoy-binbin
now we have replica-ignore-maxmemory, normally replica will be set to ignore maxmemory