Demo: believable fleet data #40

Closed
opened 2026-11-16 05:21:18 +00:00 by jcoffey-dev · 1 comment
Owner

Things that give the simulator away:

  • every filesystem on every host is 33% full;
  • hosts are fleet-0000… with no role;
  • fleet-0041 is a Redis server, a KVM host, the IBM i collector and the HMC reader at once, so IBM i alerts say "(fleet-0041)" and read as if PROD2 runs on it;
  • round numbers ("30000 of 100000 edge hits");
  • RHEL hosts list ssh.service (RHEL's unit is sshd.service).

Do: varied disks, role-based names (web-, app-, db-, kvm-, k8s-), dedicated collector hosts for storage, IBM i/HMC and virtualization, jittered counts, distro-correct unit names.

Things that give the simulator away: - every filesystem on every host is 33% full; - hosts are `fleet-0000`… with no role; - fleet-0041 is a Redis server, a KVM host, the IBM i collector and the HMC reader at once, so IBM i alerts say "(fleet-0041)" and read as if PROD2 runs on it; - round numbers ("30000 of 100000 edge hits"); - RHEL hosts list `ssh.service` (RHEL's unit is `sshd.service`). **Do:** varied disks, role-based names (web-, app-, db-, kvm-, k8s-), dedicated collector hosts for storage, IBM i/HMC and virtualization, jittered counts, distro-correct unit names.
jcoffey-dev added the
priority
p2
enhancement
labels 2026-11-16 05:21:18 +00:00
jcoffey-dev self-assigned this 2026-11-16 05:21:18 +00:00
Author
Owner

Done in #54. Demo hosts are named for their site and role (phx1-app05, phx1-db03, iad1-lb02, lon1-lb01) instead of fleet-NNNN. Storage, IBM i and the hypervisor managers are read by collector hosts of their own (phx1-stor01, phx1-ibmi01, phx1-virt01), so an IBM i alert no longer reads as if PROD2 ran on a cache server. Only app servers run libvirt, RHEL hosts run sshd and chronyd, and Akamai CP codes show traffic, not round numbers. The public demo picks this up on its next deploy.

Done in #54. Demo hosts are named for their site and role (phx1-app05, phx1-db03, iad1-lb02, lon1-lb01) instead of fleet-NNNN. Storage, IBM i and the hypervisor managers are read by collector hosts of their own (phx1-stor01, phx1-ibmi01, phx1-virt01), so an IBM i alert no longer reads as if PROD2 ran on a cache server. Only app servers run libvirt, RHEL hosts run sshd and chronyd, and Akamai CP codes show traffic, not round numbers. The public demo picks this up on its next deploy.
Sign in to join this conversation.