Skip to content

Instantly share code, notes, and snippets.

@cfebs

cfebs/rpc.md

Created July 7, 2017 23:05
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save cfebs/98a48e97cc0c7bf787f937cb78dd8e2b to your computer and use it in GitHub Desktop.
Save cfebs/98a48e97cc0c7bf787f937cb78dd8e2b to your computer and use it in GitHub Desktop.
Download ZIP
Raw
rpc.md

This is to track moving completely off of vms for rpcs.

We currently have 4 rpc boxes: rpcb1..4 and 16 vms: rpc1..16

At peak we are handling about 8k connections per second.

1 physical machine at peak is handling about 1.3k connections per sec. This is in an underutilized state too, plenty of resource room (at peak with php7 we're talking < 20% CPU and 130Mbps) 1 vm is handling about 230 connections per sec

To maintain our current capacity we should determine how many more new hardware boxes we should add accounting for the fact that rpcb1,2 need to go away.

In our current underutilized but stable configuration, if we are talking pure core equivalents:

  • VMs provide 60 cores (4 cores per 16 vms)
  • HW provide 168 (rpcb1=88, rpcb3=32, rpcb4=60)

So this means to find roughly 140 cores

With our current hardware/setup we can easily take vms away one by one to see if we hit any limits with our current hardware that would inform how much we can

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment