Merge pull request #3 from sown/tds/document-things

document dns, copy pxe docs from old wiki

docs/infrastructure/dns.md

@@ -0,0 +1,22 @@
+# DNS
+
+We have three way split DNS - an external zone that's visible to the world, a university-internal zone that's only visible inside the university network, and a SOWN-internal zone that's only visible inside SOWN.
+
+## External and universiversity-internal DNS
+
+Our domains `suws.org.uk` and `sown.org.uk` both have DNS hosted by the University, with DNS managed through their Infoblox system. Certain SOWN members have access to the web interface to update the records.
+
+The university-internal zone is also managed through this.
+
+## SOWN Internal DNS
+
+Our internal DNS for `sown.org.uk` is hosted on our legacy server `auth2` running BIND. This also hosts reverse DNS for `10.5.0.0/16` and `2001:630:d0:f700::/56`. The DNS zone is built hourly by `/etc/cron.hourly/updatednszones`.
+
+Parts of the zones are built from the legacy admin system, `node_control`, which is invoked through a PHP script and writes out temporary zonefiles. This is what generates the DNS records used for our nodes.
+
+These are combined with newer parts of the zonefile generated from Netbox. This uses netbox export templates which generate the zonefile. This is what generates the DNS records for our servers and infrastructure.
+
+The script then concatenates the zonefiles together, so the final zone is the combination of these two.
+
+## Resolvers
+Servers within SOWN should use our floating gateway addresses (`10.5.0.254` and `2001:630:d0:f700::254`) as DNS resolvers. These run BIND, and also hold our internal zones, AXFR'd from auth2. This means our internal DNS still works when auth2 is down.

docs/infrastructure/management/pxe.md

@@ -0,0 +1,24 @@
+# PXE
+PXE is a standard for booting machines over the network. In SOWN, this is used to allow servers to be remotely recovered or reinstalled - and avoiding needing to go to campus with a USB stick!
+
+On boot, the NIC's PXE ROM will get an address via DHCP, then download via TFTP and boot iPXE. iPXE then chainloads a small script (sown.ipxe) which is used to fetch a kernel and initrd over HTTP.
+
+The DHCP, TFTP and HTTP servers run on both of our GW servers. See the [ansible role](https://github.com/sown/ansible/tree/main/roles/pxe) for details. This also downloads an Ubuntu ISO and extracts the parts of the ISO needed for PXE boot.
+
+## How?
+See [our IPMI/iDRAC docs](idrac.md) for how to remotely get a console on servers.
+
+During boot, do `<Escape><@>` repeatedly to get a machine to PXE boot.
+
+## Building iPXE
+We build our iPXE like:
+```
+apt install git build-essential liblzma-dev
+git clone https://git.ipxe.org/ipxe.git
+cd ipxe/src
+echo "#define DIGEST_CMD" > config/local/general.h # enable md5sum+sha1sum
+echo "#define CONSOLE_SERIAL" > config/local/console.h # enable serial console
+make bin/undionly.kpxe
+```
+
+The DIGEST_CMD bit isn't needed now, we used to use it for validating image checksums by hand (when booting off public mirror servers).