Currently if bootstrap fails in the middle, e.g. due to using default timeout (which is 0.5 sec) or connection problems, user has no other way to retry bootstrap, since router supposes, that bootstrap ended: it's able to find buckets on some replicasets.
Vshard bootstrap should be atomic: either create buckets on all replicasets or do not create it anywhere. Otherwise, user has no other way to retry bootstrap other than reboostraping all shards.
R-omk
commented
1 week ago
Currently if bootstrap fails in the middle, e.g. due to using default timeout (which is 0.5 sec) or connection problems, user has no other way to retry bootstrap, since router supposes, that bootstrap ended: it's able to find buckets on some replicasets.
Vshard bootstrap should be atomic: either create buckets on all replicasets or do not create it anywhere. Otherwise, user has no other way to retry bootstrap other than reboostraping all shards.