When using clustering / index-replication, the master selects a list of peers, which will be target for the next bucket after a roll.
I'm curious how the master selects the next peers - is this selection intelligent (based upon free disk space, CPU/IO load or similar) or just based upon round-robin or random?
it would be great, if this algorithm could be tweaked so that uneven-sized peers can be utilized most efficient.
Best regards,
Karsten
