(Ported from memsql-public-chat Slack channel)
dweichel [12:55 PM]
Is there any way to “autoscale” the number of leafs/nodes in our configuration based on high CPU? We’re looking to address unplanned spikes in CPU on a dynamic basis.
would a combined approach of ADD LEAF
and REBALANCE PARTITIONS
do the trick?
seth [Oct 1st at 1:47 PM]
@dweichel
There is no “auto scaling” feature in MemSQL.
A database in MemSQL has a fixed number of database partitions. You can Add and Remove aggregators and leaf nodes in a MemSQL cluster. For this operations, consider using MemSQL Ops for the command line commands. ADD/REMOVE Leaf are low level cluster language commands.
When rebalancing partitions, you will want an even distribution of fixed database partitions. For example, if you have a database with 16 partitions and 4 leaf nodes, you can expand to an 8 leafs nodes (2 database partitions per leaf). If there is capacity, you can also downsize to 2 leaf nodes with 8 database partitions per leaf.
dweichel [1 month ago]
thanks Seth. We’ll explore this route
any expectations around how long it takes to rebalance? I know it’ll vary depending on the circumstance but curious if it’s in the seconds/minutes/hours realm