Hosts

srv01

../../_images/nuc8i3beh3.jpg

  • Intel NUC8i3BEH3

  • Intel Core i3-8109U (2-core, up to 3.6GHz)

  • 2 x 8GB DDR4-2666 RAM

  • 256GB SSD

This was the original workload-hosting server before those workloads outgrew it and moved into the lab architecture that runs on the three servers below. It functions as the hardened entry-point to the network for external traffic, especially SSH traffic.

lab01..06

../../_images/dell-3070.jpg

  • Dell OptiPlex 3070 Micro

  • Intel Core i5-9500T (6-core, up to 2.2GHz)

  • 2x 16GB DDR4 RAM

  • 256GB NVMe storage (OS and software)

  • 960GB or 800GB SSD storage (Ceph)

Lab servers.

These were given to us after being decommissioned as thin-clients at work. They’re not high-performance machines but they don’t need to be - they’re relatively low-power devices that are happy to run light-to-medium workloads 24/7.

nfs01

../../_images/hp-proliant-n40l.jpg

  • HP ProLiant N40L

  • AMD Turion II Neo N40L (2-core, 1.5GHz)

  • 2 x 4GB DDR3-1333 RAM

  • 500GB local HDD

  • 4TB NAS capacity

This is the network accessible storage (NAS) server in the garage. It provides bulk storage for things like CCTV recordings and backups. It doesn’t have a lot of compute power, but it has been given a lot of storage!

The bulk storage is spread across 4 x 2TB disks and has a total usable capacity of around 4TB. The total is lower than the sum of the disks because we sacrifice some space for fault-tolerance: the disks are organised into a 3-disk RAID5 array with one disk as a hot spare. Normally RAID5 can tolerate one disk failing, so having RAID5 and a hot spare means that one disk can fail, the spare disk will automatically join the array, then we still have fault tolerance while the dead disk is replaced. Rebuilding a RAID5 array also puts significant stress on the disks - having the spare reduces the likelihood of a cascade of disk failures when one fails, then rebuilding causes another to fail (which would be irrecoverable).

The RAID array is implemented in software by mdadm (as opposed to a dedicated hardware RAID controller) - you can see the RAID status by running mdadm --detail /dev/md0 as root. The RAID status is monitored as part of the observability stack, so we will get alerts if any disks fail.

desktop01

../../_images/nuc10i7fnh.jpg

  • Intel NUC10i7FNH

  • Intel Core i7-10710U (6-core, up to 4.7GHz)

  • 32 GB RAM

  • 500GB NVMe M.2 storage

Mark’s desktop.

laptop01

../../_images/framework13.jpg

  • Framework 13, 13th generation

  • Intel Core i5-1340P (4+8-core, up to 4.6GHz)

  • 64 GB RAM

  • 1TB NVMe M.2 storage

Mark’s laptop.

rpi01 and rpi02

../../_images/rpi-2b.jpg

  • Raspberry Pi 2 Model B

  • ARM Cortex-A7 CPU (4-core, 0.9GHz)

  • 1GB RAM

  • 32GB SD card storage

Small form factor and low power devices, mostly used for highly-available Pihole.