Skip to content

Server Modernization


Phase I

Phase one of the server modernization shifted away from multipurposed servers and kvms to lxc/lxd based containers. * Moving all legacy system functions onto separate linux containers isolated from each other. * Use mirrored disk systems to insure that disk corruption does not lead to data corruption. * Start giving a shit about the systems, code, and sites on them. * Own your code/data. (If your free code hosting system is shutdown or taken over by Microsoft is it really free)

Server Modernization Phase II

Phase two extends on this by integrate Ansible into system maintenance tasks.

  • Integrate Ansible into system maintenance tasks
  • Reevaluate Centos and other RPM based containers built using playbooks vs profiles/scripts/cloud-init while maintaining current security model
  • Develop off site backup strategy.
  • Clean up the cruft (If it doesn't bring you joy DTMFA)

SMP III Make Shit Happen / Own Your Shit

  • Work on secure and efficient traffic in and out of home lans (Privoxy,DNS based ad blocking,squid etc)
  • Continue to refine server operation/maintanance.
  • Build Gitlab and other alternatives to trac/git and evaluate workflows.
  • Deploy off site backup strategy.
  • Build out content.
  • Start new projects.
  • Distribute data and backups over the network to home servers.
  • Document home server/network setup


  • Security
  • Flexibility
  • Simplification


  • network
  • performance
  • disk


At present the environment contains a vpn capable router (Knight) and two enterrise class servers

  • bs2020 , a Dell PowerEdge R610 [[br]]and
  • kb2018 a HP ProLiant DL380 (g7) .


The network is divided into 3 segments

  • a private administrative lan
  • tbd.tbd.tbd.tbd/? a private vpn for home offices
  • A public facing lan.

The hosts themselves do not have any public facing interfaces and are only accessible though the admin lan. The containers which handle all public facing work do so via an anonymous bridge configuration, allowing them to access the internet directly without allowing external access to the servers.

bs2020 ports
port Interface IP Address/mask linux device purpose
1 eno1 eno1 internal / admin lan
2 ? ?.?.?.?/?? eno2 vpn for home/office networks
3 br1 eno3 Public Interface for infrastructure servers
4 br0 eno4 Public Interface for dev/deploymant servers
idrac remote console
As Drawn As Deployed.
kb2018 ports
port Interface IP Address/mask linux device purpose
4 enp4s0f1 enp4s0f1 internal / admin lan
3 enp4s0f0 ?.?.?.?/?? enp4s0f0 vpn for home/office networks
2 br1 enp3s0f1 Public Interface for infrastructure servers
1 br0 enp3s0f0 Public Interface for dev/deploymant servers
ilo remote console
As Drawn As Deployed.

See: ‚Äč which is built referencing a google doc with proposed allocations

Server OS, Filesystems and Disk layout

The servers are both running a standard install Ubuntu Server LTS, along with the Canonical supported LXD "snap". Outside of zfs not much is added to the stock installation. This is intentional. Since the real work is done by the containers the host os is considered disposable and can be rebuilt without effecting production.

Disk Layout

The system disks on both servers use hardware raid 1+0 mirroring. The containers are able to take advantage of zfs mirroring and caching.

bs2020 disks
disk device/pool bay type mount point(s) purpose/notes
Host Machine Disks
sdg /dev/sdg 0 ext4 / root filesytem (hardware raid)
sdg /dev/sdg 1 ext4 / mirror
sda1 /dev/sda1 external ext4 /archive backup staging
development zfs pool
sdc devel 2 zfs /var/lib/lxd/storage-pools/devel dev/deployment (www,trac,usw)
sdd devel 3 zfs mirror
development zfs pool
sdd infra 4 zfs /var/lib/lxd/storage-pools/infra infrastructure (email,dns,usw)
sde infra 5 zfs mirror

On kb2018 the second pair of disks are Solid State. The first partition on each is a mirrored pair for the infrastructure zfs pool. The remaining partitions are for zfs caching.

kb2018 disks
disk device/pool bay type mount point(s) purpose/notes
Host Machine Disks
sda /dev/sda 0 ext4 / root filesytem (hardware raid)
sda /dev/sda 1 ext4 / mirror
infrastructure zfs pool
sdb1 infra 2 zfs /var/lib/lxd/storage-pools/infra infrastructure (email,dns,usw)
sdc1 infra 3 zfs mirror
development zfs pool
sdd devel 4 zfs /var/lib/lxd/storage-pools/devel dev/deployment (www,trac,usw)
sde devel 5 zfs mirror
sdb2 devel 2 zfs zfs cache (proposed)

Hardware raid on the DL380

The raid controller on the Dell allows a mixing of hardware raid and direct hot swappable connections. The HP 420i does only hardware raid or direct connections (HBA) but not both. Since we use the hardware raid the remaining disks need to be configured using the ssacli or the raid controllers bios. See: DudeWhereAreMyDisks


Work previously done by standalone servers is now done though LXD managed containers. [#fn1 (1)] An up to date list of containers is maintained at''


Ansible is used to make most tasks reasonable including. * creating containers * updating containers * updating admin passwords and ssh keys. * accessing

Tasks: Accessing Hosts

bs2020/kb2020 ssh access

The host machines for the containers can be accessed through the admin lan. Currently this is done through ssh redirection. Eventually it will require a vpn connection. Only ssh key access is allowed and root is not allowed to login. To escalate privileges requires sudo.

Current ssh port mappings to

port destination
22 bs2020 ssh via admin lan
222 bs2020 racadm / serial console via ssh
2222 knight / vpn
22222 kb2018 ssh via admin lan
22223 kb2018 hpILO / serial console via ssh

note: as of a few updates ago you have to tell apples ssh client to use ssh-dss as below

steve:~ don$ ssh -p22223 -oHostKeyAlgorithms=+ssh-dss
User:feurig logged-in to / FE80::9E8E:99FF:FE0C:BAD8)
iLO 3 Advanced for BladeSystem 1.88 at  Jul 13 2016
Server Name: kb2018
Server Power: On

hpiLO-> vsp

Virtual Serial Port Active: COM2

Starting virtual serial port.
Press 'ESC (' to return to the CLI Session.

Ubuntu 18.04.1 LTS kb2018 ttyS1

kb2018 login: <ESC> (
hpiLO-> exit
steve:~ don$ ssh -p 222
/admin1-> console com2
Connected to Serial Device 2. To end type: ^\

Ubuntu 18.04.1 LTS bs2020 ttyS1

bs2020 login: <CTL> \
/admin1-> exit
CLP Session terminated
Connection to closed.
steve:~ don$

_ if the serial port is still in use do the following _

Virtual Serial Port is currently in use by another session.
hpiLO-> stop /system1/oemhp_vsp1

bs2020/kb2018 graphical console access

bs2020 allows complete control of the system via a Dell Idrac 6 controller. This also requires access to the admin lan. This is described on the [wiki:NotesOnIdrac6 Idrac 6 page] kb2020 allows similar using the on board described on the [wiki:NotesOnILO3 ILO 3 Notes page.]

ssh access to containers

The susdev profile adds ssh keys and sudo passwords for admin users allowing direct ssh access to the container.

steve:~ don$ ssh

The containers can be accessed directly from the lxc/lxd host as root

root@bs2020:~# lxc exec harvey bash
root@harvey:~# apt-get update&&apt-get -y dist-upgrade&& apt-get -y autoremove

Updating dns

Dns is provided by bind , The zone files have been consolidated into a single directory under /etc/bind/zones on naomi (

root@naomi:/etc/bind/zones# nano suspectdevices.hosts
@               IN      SOA (
                2018080300 10800 3600 3600000 86400 )
;               ^^ update ^^
; .... make some changes ....
morgan          IN      A
git             IN      CNAME   morgan
root@naomi:/etc/bind/zones# service bind9 reload
root@naomi:/etc/bind/zones# tail /var/log/messages
Sep  3 08:10:04 naomi named[178]: zone loaded serial 2018080300
Sep  3 08:10:04 naomi named[178]: zone sending notifies (serial 2018080300)
Sep  3 08:10:04 naomi named[178]: client ( transfer of '': AXFR-style IXFR started (serial 2018080300)
Sep  3 08:10:04 naomi named[178]: client ( transfer of '': AXFR-style IXFR ended
Sep  3 08:10:04 naomi named[178]: client received notify for zone ''

Updating Hosts / Containers

When updates are available Apticron sends us an email. We prefer this to autoupdating our hosts as it helps us maintain awareness of what issues are being addressed and does not stop working when there are issues. All hosts in /etc/asnsible/hosts on kb2018 shoul be updated using the following add hoc command.

feurig@kb2018:~$ sudo bash
root@kb2018:~# ansible pets -m raw -a ""

Creating containers

ansible-playbook playbooks/create-lxd-containers.yml .....YOU ARE HERE..... documenting the ansible script to create containers.

Backing Up Containers

Backing up containers using ansible is depreciated. A python script and cron tab create nightly snapshots and moves them to bs2020.

cd /etc/ansible ;screen -L time ansible-playbook playbooks/backup-lxd-containers.yml -vvv -i importants

links.... review / Cruft ..