I assume this topic has a longer history than I'm aware of. Mainly opening to discuss.
At the moment, a single-node cluster is initialized with a zone-config that defaults to 3-way replication. As a result of this, we mark all ranges in the cluster as "under-replicated".
We also log various messages that call this out, like:
E180110 04:34:11.439278 415 storage/queue.go:738 [n1,replicate] 221 replicas failing with "0 of 0 stores with all attributes matching []; likely not enough nodes in cluster"
A user can modify the zone-configs to specify a replication factor of 1, which should avoid this issue, but that it's the easiest thing to do. Is this the experience we want to require of users who want just a single-node cluster? I think a strong argument can be made for "yes", because we want to make sure any misconfigurations that prevent up-replication are called out when a multi-node cluster is desired.
I assume this topic has a longer history than I'm aware of. Mainly opening to discuss.
At the moment, a single-node cluster is initialized with a zone-config that defaults to 3-way replication. As a result of this, we mark all ranges in the cluster as "under-replicated".
We also log various messages that call this out, like:
A user can modify the zone-configs to specify a replication factor of 1, which should avoid this issue, but that it's the easiest thing to do. Is this the experience we want to require of users who want just a single-node cluster? I think a strong argument can be made for "yes", because we want to make sure any misconfigurations that prevent up-replication are called out when a multi-node cluster is desired.