Reputation: 13
How could I latency measure replication in Cassandra, NoSQL, or quantify of replicas updates in some time?
Upvotes: 0
Views: 296
Reputation: 16400
What level of precision your looking for? To a point its not possible in a distributed system given clock drift and the speed of light. Easiest would be coordinator write latency as a measure of the round trip plus a little. Could also do a ping mapping, but once again sub ms its pretty difficult. The writes are measured in microseconds so hopefully the network latency should be dominating it with exceptions of the impact of GCs and OS.
The org.apache.cassandra.metrics:type=Messaging,name=CrossNodeLatency
metric might be meaningful as a C* level measure of time it takes a message to go from one node to another (or between DCs) but it does not account for clock drift.
Upvotes: 1