> 1) If all nodes are up: > - Will all writes eventually reach all nodes (of the 3 nodes)?
I believe that if read repair is completely off, then for data that was written that did *not* get saved by hinted hand-off, would not propagate until anti-entropy as part of a 'nodetool repair' or perhaps as part of node movement in the ring (as a side-effect). Also see http://wiki.apache.org/cassandra/Operations under "Consistency". > - What will be the maximal time until the last write reaches the last node > (of the 3 nodes)? (e.g. Assume one of the node is doing compactation at that > time) There is no particular time guarantee, unless you yourself take steps that would imply such a guarantee (such as by running repair with a certain frequency). > 2) If one or two nodes are down > - As I understood it, one node will buffer the writes for the remaining > nodes. AFAIK not all. I.e., only when a node is marked as down will hinted hand-off start eating writes for the node (right, anyone?). Hinted hand-off is not supposed to be a guarantee that all data will become up-to-date; it's rather a way to lessen the impact of nodes going down by decreasing the amount of data that remains out of synch. > - If the nodes go up again: When will these writes be propagated, at > compactation?, what will be the maximal time until the writes reach the 2 > nodes? Will these writes be propagated at all? Again there's no time guarantee as such. As for the writes, I believe hinted hand-off sends those along independently of compaction (but I'm not sure). -- / Peter Schuller
