NodeSync can place higher load on your cluster than traditional repair. I have seen some clusters that were OK with OpsCenter repair service or manual repairs get overloaded with NodeSync. So I would recommend testing NodeSync out with a realistic workload to make sure your cluster can handle it, and be ready to disable it quickly if there are signs of overloading (high latencies, timeouts, or high CPU or I/O utilization). Currently NodeSync is only available for DSE, not Apache Cassandra.
The primary advantage of NodeSync is that it is operationally simpler. It uses the read repair mechanism, so there is less possibility of a hanging or long running repair, you would need to restart the whole process if a node that was streaming goes down. On Thu, Apr 2, 2020 at 10:57 AM Aakash Pandhi <pandhiaak...@yahoo.com.invalid> wrote: > Hi All, > > I am reviewing our data sync procedures to improve so need your input on > NodeSync. > > Are there any cons of implementing NodeSync over Repair? Is NodeSync a > future direction for cluster wide data sync? > > Thank You, > Aakash >
