The Aerospike Knowledge Base has moved to https://support.aerospike.com. Content on https://discuss.aerospike.com is being migrated to either https://support.aerospike.com or https://docs.aerospike.com. Maintenance on articles stored in this repository ceased on December 31st 2022 and this article may be stale. If you have any questions, please do not hesitate to raise a case via https://support.aerospike.com.
FAQ: Can Aerospike run on a heterogenous cluster?
Detail
What happens if the nodes in a cluster are not all identical in hardware or software?
Answer
There isn’t anything that stops Aerospike from starting up on a cluster where nodes have different OS, hardware, or Aerospike Server versions, but we don’t recommend running in that state in production.
-
Heterogenous OS
As long as each node has a supported OS and version, the cluster will run fine in normal operation, but it may be more difficult to configure OS parameters (different nodes needing different commands or files), and it is likely to be more difficult to troubleshoot any problems that may arise. We do not recommend running a cluster this way if you can avoid it.
-
Heterogenous Aerospike version
Aerospike can run a mixed cluster during a rolling upgrade, but features of the later version may not function correctly until all nodes have been upgraded to the same version.
-
Heterogenous Aerospike configuration
This is normally not recommended, and some settings must be the same across the cluster or the cluster won’t start up, but sometimes it may be necessary to have different configuration on some nodes. See the configuration reference for details.
-
Heterogenous hardware
Aerospike distributes data partitions, and therefore all the other kinds of load, such as client connections, evenly across the nodes (more or less). This means that if the node with the least disk space, memory, CPU cores, or any other kind of resource can’t handle 1/N of the data, the cluster as a whole suffers. For example, if a node with less memory or storage breaches the high-water mark and begins evicting records, all nodes across the cluster will perform evictions for that namespace, even if they have not breached the HWM. Looked at another way, any capacity the larger nodes have in excess of the smallest node is effectively wasted, possibly to the point where the cluster would run more effectively without the smaller nodes.
Keywords
NODE CAPACITY PLANNING CLUSTER HETEROGENOUS
Timestamp
June 13 2019