|
Size: 11499
Comment: Add --hostname=illustrious so smokeping shows correct hostname on graph titles
|
Size: 17571
Comment: add sysctl for network receive buffer
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 22: | Line 22: |
= docker containers = * elasticsearch - for UMAD development * pihole-exporter - fetches pihole stats, publishes for prometheus: https://hub.docker.com/r/ekofr/pihole-exporter * prometheus - receive stats and store them * smokeping * unifi-controller |
|
| Line 164: | Line 172: |
| === Enable IPv6 for containers === This doesn't happen out of the box, so we need to do it ourselves. Here's the official docs: https://docs.docker.com/config/daemon/ipv6/ {{{#!wiki caution '''The notes immediately below are if you don't have IPv6 connectivity to the internet''' If you have functioning IPv6, like with a delegated prefix, you'll need to choose a subnet inside that prefix. More notes below in the ndppd section. }}} Amend the config in `/etc/docker/daemon.json` like so, adding the IPv6 keys: {{{ { "log-driver": "json-file", "log-opts": { "max-size": "50m", "max-file": "10" }, "ipv6": true, "fixed-cidr-v6": "fd21:1268:04a5:d0c::/64" } }}} We're using a subnet with [[https://en.wikipedia.org/wiki/Unique_local_address| unique local addressing]]. I've carefully chosen a subnet that's unique, and ensured it's a ''good soize'' - the subnet needs to be at least /80 or larger so it can jam the 48-bit virtual MAC address into it at the end. Here's some code to generate a good prefix: {{{#!python import random h = bytearray(random.randbytes(5)).hex() print(f"fd{h[0:2]}:{h[2:6]}:{h[6:10]}:d0c::/64") fd21:1268:04a5:d0c::/64 }}} The `fc00::/7` space is recommended to be carved into /64 subnets for your site like so: * Use fd00/8 as the stem ('d'/1101 indicates locally-assigned addresses) * Choose 40 bits using high quality randomness as the unique site code * Then you have 16 bits to use as subnet IDs, I'm using `:d0c:` for obvious reasons. Then `systemctl restart docker`, and now it should Just Work I guess. Your docker0 bridge will get the additional /64 subnet on it, and your containers will get an autoassigned IPv6 address. For local-only purposes this should be sufficient, but you probably won't be able to get off the docker host and reach other machines on the segment. To do this you either need to NAT (eww), or get the host to respond on behalf of the bridge subnet for NDP requests (IPv6 equivalent of ARP). What I can see at this point is that packets go out fine, but can't make it back. * ping6 requests leave the container with nexthop = fd21:1268:4a5:d0c::1 * docker host (fd21:1268:4a5:d0c::1) gets the packet and forwards it out eno1 towards the target * Target receives ping request, sends reply to fd21:1268:4a5:d0c:0:242:ac11:2 * Reply is routed to the default gateway, and this is the sticking point now I think we need an NDP proxy... === NDP Proxy ndppd === Problem is as described here, as as above: https://forums.docker.com/t/ipv6-not-working/10171/7 And here's a guide mentioning using ndppd: https://medium.com/@skleeschulte/how-to-enable-ipv6-for-docker-containers-on-ubuntu-18-04-c68394a219a2 {{{ apt install ndppd cp /usr/share/doc/ndppd/ndppd.conf-dist /etc/ndppd.conf vim /etc/ndppd.conf }}} '''You need to select the correct subnet here'''. If you've got public routable address space like I do, you need to carve out a small subnet. My LAN is a /64 subnet, so I'm using a /80 chunk of that and giving it to docker. You'll need to enter the same /80 subnet in `/etc/docker/daemon.json` as mentioned above. Tweak up the rule for our subnet and interfaces: {{{ proxy eno1 { rule fd21:1268:04a5:d0c::/64 { # Either of these two would work fine, the default is auto auto iface docker0 } } }}} Restart ndppd and it should answer NDP queries now. Whew! All this just so that smokeping can reach out to IPv6 hosts outside the docker host. ==== Problems you might have === I couldn't get this working at first. What I determined was going on is that other machines don't even bother doing an NDP query when I try to ping anything in the docker range. I think they look at the destination, realise it couldn't possibly be in the same subnet/LAN as them, so they go via layer 3 (the router) instead of sticking layer 2. Which means we need to be in the same /64 subnet as the rest of the LAN, a sibling to the docker host but living inside the docker host. They'll be hidden behind the bridge, and ndppd will make them visible to the rest of the network. I've grabbed `2404:e80:42e3:0:d0c::/80` and that works now, other hosts will find it with NDP. The nice thing is that I could even have other docker hosts on the network and use the same config. They'll generate different virtual MACs for each container, and they'll live in the same /80 subnet. Containers on host A wouldn't be able to find containers on host B, but non-docker hosts would be able to find any docker container in that /80 subnet, because ndppd will only answer a query if it owns that container with the requested address. == Network tuning == Caddy complains that it can't set a large receive buffer for network connections: {{{ Nov 03 11:22:19 illustrious.thighhighs.top caddy[1601301]: {"level":"info","ts":1667434939.2639456,"msg":"failed to sufficiently increase receive buffer size (was: 208 kiB, wanted: 2048 kiB, got: 416 kiB). See https://github.com/lucas-clemente/quic-go/wiki/UDP-Receive-Buffer-Size for details."} }}} So let's increase it. {{{ echo -e "# Increase max recv buffer to about 2.5MiB\nnet.core.rmem_max=2500000" >> /etc/sysctl.d/20-network-recv-buffer.conf }}} |
|
| Line 371: | Line 483: |
= NFS mount from NAS = Want to sort through my files, and the NAS is well setup for this usage. 1. `apt install nfs-common` 1. Let's mount it with systemd, create a new unit for the mount: `systemctl edit --force --full cargo.mount` {{{ [Unit] Description=cargo volume from iowa After=network.target [Mount] What=iowa.thighhighs.top:/volume1/cargo Where=/cargo Type=nfs Options=_netdev [Install] WantedBy=multi-user.target }}} 1. Mount it: `systemctl start cargo.mount` |
- Running at home, general purpose server
- Intel NUC (BOXNUC7i3BNH)
- Core i3-7100U at 2.40GHz
- 500GB WD Blue SN550 NVMe M.2
- 8gb DDR4 2400MHz Kingston KVR24S17S8/8
- Ubuntu 21.10
Contents
users
- furinkan
services
- ssh
- http/s
Smokeping (available on internal IP, and IPv6)
docker containers
- elasticsearch - for UMAD development
pihole-exporter - fetches pihole stats, publishes for prometheus: https://hub.docker.com/r/ekofr/pihole-exporter
- prometheus - receive stats and store them
- smokeping
- unifi-controller
build notes
On 2022-01-02, FDE on LVM, Ubuntu server 21.10
Had some problem with the curtin unpack of the base image, I changed the Ubuntu archive URL to Datamossa in AU and hey presto it worked. Guess something was corrupted, shrug.
- Mostly default settings
- Enable full disk encryption with LVM, default settings
- Enable sshd during install
basic env
Install your authorized_keys into root and furinkan user
Enable NOPASSWD sudo
%sudo ALL=(ALL:ALL) NOPASSWD: ALL
Set hostname:
hostnamectl set-hostname illustrious.thighhighs.top
Set timezone
timedatectl set-timezone Australia/Sydney
Set editor
echo "export EDITOR=vim" > /etc/profile.d/editor-vim.sh
Disable HashKnownHosts
echo -e "Host *\n HashKnownHosts no" > /etc/ssh/ssh_config.d/99-global.conf
Configure screen
curl -o ~/.screenrc https://gist.githubusercontent.com/barneydesmond/d16c5201ed9d2280251dfca7c620bb86/raw/.screenrc
Configure top by entering this cheatcode
z x c b s 1.5 <Enter> e <zero> 1 W q
Fix locales, select en_AU.UTF-8: dpkg-reconfigure locales
Disable console blanking, seems this is already done by default: cat /sys/module/kernel/parameters/consoleblank
- Already set to zero means it shouldn't blank
Disable wifi and bluetooth, we don't need them and it slows down boot
systemctl disable wpa_supplicant.service --now systemctl disable bluetooth.target --now
Install useful packages
apt update apt install -y vim screen bash-completion lsof tcpdump netcat strace nmap less bsdmainutils tzdata whiptail netbase wget curl python-is-python3 net-tools ack jq make elinks nmap whois ethtool bind9-dnsutils apt-utils man-db plocate
Do a full upgrade, index the system for locate, then reboot
apt full-upgrade updatedb reboot
Configure networking
Use netplan for this, it's convenient and easy.
cd /etc/netplan/
mv 00-installer-config.yaml 00-installer-config.yaml.disabled
vim 10-thighhighs.yaml
network:
version: 2
ethernets:
eno1:
critical: true
dhcp-identifier: mac
dhcp4: false
dhcp4-overrides:
use-dns: false
dhcp6: true
dhcp6-overrides:
use-dns: false
ipv6-privacy: false
addresses:
- "192.168.1.12/24"
# :12 for the .1.12 IPv4
- "2404:e80:42e3:0:12:0:0:12/64"
routes:
- to: 0.0.0.0/0
via: 192.168.1.1
on-link: true
nameservers:
addresses:
- 192.168.1.20
- 192.168.1.24
- fe80::e65f:1ff:fe1c:c6ea
- fe80::ba27:ebff:fe8c:f4f8
search:
- thighhighs.topTry applying it with netplan try, see if your SSH session still works, then go ahead and reboot if it's good.
Setup clevis for automated decrypt on boot
- apt install clevis-luks
Bind the volume to the Tang servers
clevis luks bind -d /dev/nvme0n1p3 sss '{"t": 1, "pins": {"tang": [{"url": "http://ocular.thighhighs.top:8888"},{"url": "http://funicular.thighhighs.top:8888"}]}}'- apt install clevis-initramfs
Test by rebooting.
Docker Engine for services
Run with the official docs: https://docs.docker.com/engine/install/ubuntu/#install-using-the-repository
Prep repo
apt install ca-certificates curl gnupg lsb-release curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable" > /etc/apt/sources.list.d/docker.list
Install packages
apt update apt install docker-ce docker-ce-cli containerd.io
Test that it's working
docker run hello-world
Setup log rotation in /etc/docker/daemon.json
{
"log-driver": "json-file",
"log-opts": {
"max-size": "50m",
"max-file": "10"
}
}
Prepare space for volumes
We'll use this later for dockerised apps.
mkdir /data
Enable IPv6 for containers
This doesn't happen out of the box, so we need to do it ourselves.
Here's the official docs: https://docs.docker.com/config/daemon/ipv6/
The notes immediately below are if you don't have IPv6 connectivity to the internet
If you have functioning IPv6, like with a delegated prefix, you'll need to choose a subnet inside that prefix. More notes below in the ndppd section.
Amend the config in /etc/docker/daemon.json like so, adding the IPv6 keys:
{
"log-driver": "json-file",
"log-opts": {
"max-size": "50m",
"max-file": "10"
},
"ipv6": true,
"fixed-cidr-v6": "fd21:1268:04a5:d0c::/64"
}We're using a subnet with unique local addressing. I've carefully chosen a subnet that's unique, and ensured it's a good soize - the subnet needs to be at least /80 or larger so it can jam the 48-bit virtual MAC address into it at the end.
Here's some code to generate a good prefix:
The fc00::/7 space is recommended to be carved into /64 subnets for your site like so:
- Use fd00/8 as the stem ('d'/1101 indicates locally-assigned addresses)
- Choose 40 bits using high quality randomness as the unique site code
Then you have 16 bits to use as subnet IDs, I'm using :d0c: for obvious reasons.
Then systemctl restart docker, and now it should Just Work I guess. Your docker0 bridge will get the additional /64 subnet on it, and your containers will get an autoassigned IPv6 address.
For local-only purposes this should be sufficient, but you probably won't be able to get off the docker host and reach other machines on the segment. To do this you either need to NAT (eww), or get the host to respond on behalf of the bridge subnet for NDP requests (IPv6 equivalent of ARP).
What I can see at this point is that packets go out fine, but can't make it back.
- ping6 requests leave the container with nexthop = fd21:1268:4a5:d0c::1
- docker host (fd21:1268:4a5:d0c::1) gets the packet and forwards it out eno1 towards the target
- Target receives ping request, sends reply to fd21:1268:4a5:d0c:0:242:ac11:2
- Reply is routed to the default gateway, and this is the sticking point now
I think we need an NDP proxy...
NDP Proxy ndppd
Problem is as described here, as as above: https://forums.docker.com/t/ipv6-not-working/10171/7
And here's a guide mentioning using ndppd: https://medium.com/@skleeschulte/how-to-enable-ipv6-for-docker-containers-on-ubuntu-18-04-c68394a219a2
apt install ndppd cp /usr/share/doc/ndppd/ndppd.conf-dist /etc/ndppd.conf vim /etc/ndppd.conf
You need to select the correct subnet here. If you've got public routable address space like I do, you need to carve out a small subnet. My LAN is a /64 subnet, so I'm using a /80 chunk of that and giving it to docker. You'll need to enter the same /80 subnet in /etc/docker/daemon.json as mentioned above.
Tweak up the rule for our subnet and interfaces:
proxy eno1 {
rule fd21:1268:04a5:d0c::/64 {
# Either of these two would work fine, the default is auto
auto
iface docker0
}
}Restart ndppd and it should answer NDP queries now.
Whew! All this just so that smokeping can reach out to IPv6 hosts outside the docker host.
==== Problems you might have ===
I couldn't get this working at first. What I determined was going on is that other machines don't even bother doing an NDP query when I try to ping anything in the docker range. I think they look at the destination, realise it couldn't possibly be in the same subnet/LAN as them, so they go via layer 3 (the router) instead of sticking layer 2.
Which means we need to be in the same /64 subnet as the rest of the LAN, a sibling to the docker host but living inside the docker host. They'll be hidden behind the bridge, and ndppd will make them visible to the rest of the network. I've grabbed 2404:e80:42e3:0:d0c::/80 and that works now, other hosts will find it with NDP.
The nice thing is that I could even have other docker hosts on the network and use the same config. They'll generate different virtual MACs for each container, and they'll live in the same /80 subnet. Containers on host A wouldn't be able to find containers on host B, but non-docker hosts would be able to find any docker container in that /80 subnet, because ndppd will only answer a query if it owns that container with the requested address.
Network tuning
Caddy complains that it can't set a large receive buffer for network connections:
Nov 03 11:22:19 illustrious.thighhighs.top caddy[1601301]: {"level":"info","ts":1667434939.2639456,"msg":"failed to sufficiently increase receive buffer size (was: 208 kiB, wanted: 2048 kiB, got: 416 kiB). See https://github.com/lucas-clemente/quic-go/wiki/UDP-Receive-Buffer-Size for details."}So let's increase it.
echo -e "# Increase max recv buffer to about 2.5MiB\nnet.core.rmem_max=2500000" >> /etc/sysctl.d/20-network-recv-buffer.conf
Reverse proxy for services
Proxy software
Let's try out Caddy, I've been curious for a while now and it might meet all my needs.
Use official docs for a repo-packaged version: https://caddyserver.com/docs/install
apt install -y debian-keyring debian-archive-keyring apt-transport-https curl -1sLf https://dl.cloudsmith.io/public/caddy/stable/gpg.key > /etc/apt/trusted.gpg.d/caddy-stable.asc curl -1sLf https://dl.cloudsmith.io/public/caddy/stable/debian.deb.txt > /etc/apt/sources.list.d/caddy-stable.list apt update apt install caddy
This uses a systemwide config in /etc/caddy/Caddyfile, and acts as a generic HTTP server initially. It's serving up a Caddy landing page from /usr/share/caddy at http://illustrious.thighhighs.top/
SSL cert
Pop it in /etc/ssl like usual.
cd /etc/ssl/ # This is a one-time action openssl dhparam -out dhparams.pem 4096 # Then copy the cert and key and intermediate CA chain here cp KEY CERT /etc/ssl/ chgrp caddy /etc/ssl/STAR_*
Smokeping
Run using the docker container, it's more convenient and separates config+data from the installation.
https://hub.docker.com/r/linuxserver/smokeping
Prepare space for data and config using a logical volume
lvcreate -L 1G -n smokeping ubuntu-vg mkfs.ext4 /dev/ubuntu-vg/smokeping mkdir /data/smokeping # Add to fstab # Smokeping config and data /dev/disk/by-uuid/a40142d8-06e0-44d7-b8bc-a3e20662cde2 /data/smokeping ext4 defaults 0 1 mount /data/smokeping mkdir /data/smokeping/config mkdir /data/smokeping/data chown -R 1000:1000 /data/smokeping/config /data/smokeping/data
Run the container:
docker run -d \ --name=smokeping \ --hostname=illustrious \ -e PUID=1000 \ -e PGID=1000 \ -e TZ=Australia/Sydney \ -p 127.0.0.1:8000:80 \ -v /data/smokeping/config:/config \ -v /data/smokeping/data:/data \ --restart unless-stopped \ lscr.io/linuxserver/smokeping
Map it through with some caddy config
smokeping.thighhighs.top {
reverse_proxy localhost:8000
tls /etc/ssl/STAR_thighhighs_top.crtbundled /etc/ssl/STAR_thighhighs_top.key
}Reload the config, and you should have a working smokeping again! \o/
systemctl reload caddy.service
Unifi controller
Move it to a slightly beefier machine, running it via Docker for cleanliness. Their insistence on only supporting Java 8 runtime is a nightmare, but perfect for Docker abstraction.
Prepare space
This is a somewhat bigger system that needs more diskspace
lvcreate -L 4G -n unifi ubuntu-vg mkfs.ext4 /dev/ubuntu-vg/unifi mkdir /data/unifi ## Add to fstab, use blkid to find the UUID # Unifi controller data /dev/disk/by-uuid/0a13b90e-904a-4803-896f-0f82e4a36518 /data/unifi ext4 defaults 0 1 mount /data/unifi mkdir /data/unifi/config chown -R 1000:1000 /data/unifi/config
Run container
docker run -d \ --name=unifi-controller \ -e PUID=1000 \ -e PGID=1000 \ -e MEM_LIMIT=1024 `#optional` \ -e MEM_STARTUP=1024 `#optional` \ -p 192.168.1.13:3478:3478/udp \ -p 192.168.1.13:10001:10001/udp \ -p 192.168.1.13:8080:8080 \ -p 192.168.1.13:8443:8443 \ -p 192.168.1.13:1900:1900/udp `#optional` \ -p 192.168.1.13:8843:8843 `#optional` \ -p 192.168.1.13:8880:8880 `#optional` \ -p 192.168.1.13:6789:6789 `#optional` \ -p 192.168.1.13:5514:5514/udp `#optional` \ -v /data/unifi/config:/config \ --restart unless-stopped \ lscr.io/linuxserver/unifi-controller
Migration
Import the backup from old controller, then on the old controller switch the inform URL to the new IP address. We'll fix up DNS afterwards.
Set the inform URL again on the new controller, because you've just restored a backup with the old IP.
TLS cert for unifi
Faff with the keystore so you can jam in your publicly signed cert. This is a script that I found and adapted.
# From https://util.wifi.gl/unifi-import-cert.sh which is now dead
# Modified by Barney Desmond on 2021-04-20 to just use a normal static paid-for cert.
# Author: Frank Gabriel, 01.01.2019
# Credits Kalle Lilja, @SprockTech and others
# Script location: /etc/letsencrypt/renewal-hooks/post/unifi-import-cert.sh (important for auto renewal)
# Tested with Debian 9 and UniFi 5.8.28, 5.9.22 and 5.9.32 - should work with any recent Unifi and Ubuntu/Debian releases
# This is the host-side of a Docker volume, you need to run this inside the
# container unless you have keytool installed outside.
UNIFI_DATADIR=/data/unifi/config/data
# XXX: this is the path inside the container
UNIFI_DATADIR=/config/data
# Backup previous keystore
cp -a "${UNIFI_DATADIR}/keystore" "${UNIFI_DATADIR}/keystore.backup.$(date +%F_%R)"
#cp -a /var/lib/unifi/keystore /var/lib/unifi/keystore.backup.$(date +%F_%R)
# XXX: do this manually on the host, outside of the container.
# Convert cert to PKCS12 format
# Ignore warnings
#openssl pkcs12 -export \
# -inkey /etc/ssl/STAR_thighhighs_top.key \
# -in /etc/ssl/STAR_thighhighs_top.crt \
# -out /etc/ssl/STAR_thighhighs_top.p12 \
# -name unifi -password pass:unifi
# Install certificate
# Ignore warnings
keytool -importkeystore \
-deststorepass aircontrolenterprise \
-destkeypass aircontrolenterprise \
-destkeystore "${UNIFI_DATADIR}/keystore" \
-srckeystore STAR_thighhighs_top.p12 \
-srcstoretype PKCS12 \
-srcstorepass unifi \
-alias unifi \
-noprompt
# Restart UniFi controller
#systemctl restart unifi
Web space for thighhighs domain
Create LV for data
lvcreate -L 1G -n www ubuntu-vg mkfs.ext4 /dev/ubuntu-vg/www mkdir /data/www ### Add to fstab # webdir /dev/disk/by-uuid/a40142d8-06e0-44d7-b8bc-a3e20662cde2 /data/www ext4 defaults 0 1 mount /data/www mkdir /data/www/illustrious chown -R furinkan. /data/www/illustrious
- Throw some content in there
Add a stanza to /etc/caddy/Caddyfile
*.thighhighs.top { root * /data/www/illustrious file_server tls /etc/ssl/STAR_thighhighs_top.crtbundled /etc/ssl/STAR_thighhighs_top.key }Reload the config: systemctl reload caddy
NFS mount from NAS
Want to sort through my files, and the NAS is well setup for this usage.
apt install nfs-common
Let's mount it with systemd, create a new unit for the mount: systemctl edit --force --full cargo.mount
[Unit] Description=cargo volume from iowa After=network.target [Mount] What=iowa.thighhighs.top:/volume1/cargo Where=/cargo Type=nfs Options=_netdev [Install] WantedBy=multi-user.target
Mount it: systemctl start cargo.mount