ClusterControl Hardware Requirements

Comments

3 comments

  • Avatar
    Paul Namuag

    Hi William Norris,

     

    That's a great question indeed. Even with a smaller ram of 2GiB (1GiB), based on my tested env, it rans smoothly for managing a 3-node cluster with HaProxy and Prometheus running on a shared virtualized host. However, if you have 100 or 1000 nodes, this really depends on how and what do you want to deploy for your CC. CC can deploy load balancers and that's up to you if you run on the same host/server where CC resides.  To be particular with your question, given that it can handle 1GiB let say a 5-8 nodes (given that the setup is dedicated to CC itself on a given host i.e. without load balancers and other servers), then running 100 to 1000 might have be good start to have >10 GiB of memory and > 120 GiB of memory respectively. For CPU, it should able to perform well at > 12 up or > 20 cores for handling 1000 nodes. But then again, this really depends though on what type of setup does the CC host is intended. 

     

    Also CC can run on a multi-tenant or run in a virtualized or containerized instance. 

  • Avatar
    William Norris

    Thanks Paul.

    Have you seen any recommendations on if the resources should be dedicated if you are running CC community on an enterprise scale.  So would only be using the monitoring & advisors, database deployment and automated performance advisors.

  • Avatar
    Paul Namuag

    Hi William,

    My take on this is it depends on your setup but running on a shared or multi-tenant host which could be easier to scale up or down would be much easier and cost efficient. It would be expensive if running on a dedicated host if you are about scalability and high availability for your ClusterControl host (as per 1.7.2 of ClusterControl –under beta as of this time or see https://support.severalnines.com/hc/en-us/articles/212425943-ChangeLog). 

     

Please sign in to leave a comment.

Powered by Zendesk