I wouldn't say it's not possible. It's just not very straightforward. The needs of each cluster will be different. For instance, some clusters will need to scale based on disk usage and ingestion rate, while others will need to scale based on latency and throughput. A p90 SLA looks a lot different than a p99.
Auto scaling down a cluster always sounds odd to me. Those use cases seem like they're in response to read heavy workloads that have calmed down. Usually you'll get better bang for the buck by using a cheap caching layer that can get orders of magnitude more throughput & be launched in minutes instead of the hours or days it takes to bring up a new node (depending on density & failures). On Thu, Mar 9, 2017 at 10:05 AM Matija Gobec <matija0...@gmail.com> wrote: > Hi, > > Autoscaling is not possible with a Cassandra cluster. Any topology change > triggers series of streaming and data shuffle around the cluster. Scaling > the cluster up or down is an operational challenge which is usually planned > in production because of the performance impact it can make. > > Matija > > On Thu, Mar 9, 2017 at 6:04 PM, D. Salvatore <dd.salvat...@gmail.com> > wrote: > > Hi, > I am using Cassandra 3 with a single DC. I would like to know if there is > any tool available for scaling up and down Cassandra automatically. > > Thanks > > >
