This post is basically what we said in the state of the network talk - just in written form with some ideas added.

It is intended to be something we return to, when we start the planning for BH2020.

Redundancy all the way

We had a couple of single-points of failure in the 2019 setup.

  • the uplink fiber (well, probably not much to do about that)
  • the stateful router (vSRX), later physical SRX240H2
  • the main router (vMX)
  • Single fiber to both TLK and BLD
  • Core Aristas (Simple config and physical connections)

We’ll make contingency plan for non-redundant parts.

The direct passive link from R31 to the GC well eliminated the need for a UPS in POP-FML.

Not both vSRX and vMX on the same box

Same as the above. Could we get 2x VEP?

Don’t rely on off-site measurements

This is the #tooshort issue. At BH2020, we will reuse the basic layout with fibers.

The map has been updated with the implemented fibers (and length), so this is a non-issue for BH2020.

Network planned up 24H gave time to solve upstream issues

Yes, we plan for having the network up 24H in advance. We should prioritize (like we did this year) to have some sort of upstream test up as soon as possible.

An easy reporting/request option for users

More coordination with our users would be nice. Using the IRC channel works for some people, but some webinterface/twitter/whatever would be useful for “tasks” and reports of problems.

The NOC DECT phone came online late. We used radios instead. It worked, but DECT is easier for “common” users.

POP improvements needed

We have been discussing datenklos. Some like it, some don’t.

The solution that we had this year with the pop boxes work, and with a better tarp solution, it will work in the sun also.

If we borrow a lot of expensive equipment the supplier might have opinions on locks, in-box climate and so on.

We have set aside the schuko->danish plugs. We will need more, unless we are able to get a lot of hig temperature cables.

#tooshort also for user cables

Distances are long at the venue. So either people bring longer cables or we add secondary POPs.

We could find a sponsor for long cables, or we could find fanless switches that would be simple install in peoples tents.

Onsite wifi guy

John the wifi guy was onsite 2-3 days, setting up, taking down the wireless and an extra day for adding some more APs.

It would be nice to have an onsite wifi guy, that could nurse the wifi and handle smaller issues continuously.

This could be John or it could someone from the actual NOC team doing it in collaboration with John.

Revision and version control

We want some sort of version control and configuration environment where we don’t log into the switches/routers.

Having the config locked down in advance and then use e.g. ansible to do updates of interfaces, passwords and such would be a prefered solution.

Serverroom preparation

Servers for (reverse-)DNS, DHCP, grafana, logs, monitoring, librenms, radius, jump host and so on can be mostly configured in advance. This would give us a drop-in functioning serverroom.

This would include having defined subnets and ranges in advance.

Servers

We would like to give people the option of placing a server in the serverroom and give them a sstic IP.

We still want a UPS for servers, and connecting them to something else than the plug next to the door would be prefered.