ETCD Recommendations for Platform9 Stack

Problem

This article talks about resource recommendations for ETCD component in the Kubernetes clusters deployed on Platform9 Managed Kubernetes (PMK) or Platform9 Self Managed Cloud Platform (SMCP) environments.

Please Note: The recommendations suggested here come from a broader perspective of known issues and the recommendations on the Official ETCD Documentations and should be not considered as an absolute hardware recommendation for optimal ETCD performance. The performance of ETCD has numerous contributing factors and in cases of issues, would often need expert troubleshooting.

However, these can certainly be treated as minimum recommendations for an optimal ETCD performance.

Environment

• Platform9 Managed Kubernetes - v4.0 and Higher
• Platform9 Self Managed Cloud Platform - v5.9.1 and Higher

Answer

• For small clusters, below are the minimum resource and disk recommendations.
• 4 vCPUs, 8GB Memory
• For large clusters, below are the minimum resource and disk recommendations.
• 8 - 16 vCPUs, 16 - 64GB Memory.

Disk recommendations:

Fast backend disks are quintessential for ETCD Performance and Stability. Typically, SSDs are recommended to be used as ETCD storage.

If Spinning Disks are to be used as ETCD storage, the recommendation is to use the fastest possible disk of RPMs at around 15000