Task #6519
Updated by Nico Schottelius about 5 years ago
We need to add / change:
* CHANGE mirror.ungleich.ch cannot be used in cdist types that are used on servers
** I suggest to introduce "mirror.placeX.ungleich.ch" in each place
** I suggest to point mirror.placeX.ungleich.ch to the active router IP (IPv6 only)
** I suggest to use http[s]://mirror.placeX.ungleich.ch in the netboot image and cdist
* CREATE mirror.placeX.ungleich.ch
** Install nginx on both routers
** Ensure that the essential packages are present
*** Devuan packages
*** hwraid repo
*** consul / prometheus (?)
** Only usable for our own range (i.e. 2a0a:e5c0::/29 and the other v6 network)
** Use nftables for it
* code.ungleich.ch cannot be used in cdist types that are used on servers:
** code.ungleich.ch is a VM
** if the VM is down, servers don't get configured
** Suggestion: mirror the ungleich-tools repo to the routers, mirror.placeX..., accessible by http(s)
* We will try to connect all systems to UPS ONLY
** My theory is that because they are also connected to the regular grid they experience an outage
** We started with router1
** Need to get in touch with Juanito or Bernegger (electricity company) to test whether UPS only setup works
* Update the monitoring infrastructure:
** ensure that prometheus (port 9090) is not reachable without authentication
** Ensure that there is 1 entry point for both monitoring systems
** Ensure that changes (dashboards) are saved to both monitoring systems
** Ensure that all production systems are monitored
Objective:
* Servers only depend on routers
* If there is no network to the outside, servers are still booted/configured
* If VMs are down, server are still booted/configured
Jason, can you coordinate this with Dominique (consultant), Jin-Guk (implementation) and Roli/Marc/Sami (learning, understanding)?