I was having compatibility issues with some NICs on my desktop and wanted to be able to quickly test them without rebooting. PCIe hotplug can be a pain to get working on hardware that doesn’t natively support it – there are a couple of good blog posts from a different Matt that you should read (part 1, part 2). Those blog posts focus on getting Thunderbolt working on a machine that doesn’t natively support it, but a lot of it is still relevant here, and Thunderbolt will actually come into the picture here in a roundabout way. I will warn you that there are caveats here, but read on if you want to see how to do this.

RouterOS 7.17 added a very nice manageability feature which is the ability to use interface lists in Bridge VLAN definitions. Between this and automatic entries based on PVIDs, you can greatly simplify configurations from the pre-7.17 world.

So far, I’ve been using nothing but RAPL and my servers’ PMBus support to measure power usage. The problem is, this means I can’t measure the rest of my rack – only a couple servers. I’ve still got two switches, a mini PC, and a cable modem. Plus, the router host does not have PMBus, so I don’t have any measure of how much power it consumes outside of the CPU and memory.

The solution to this is a metered-by-outlet PDU. The problem is that these are expensive – certainly to the point where it is never going to pay for itself in a timely manner. Buying one secondhand is the only economically viable way to do this, and it’s going to be a very old unit.

The original Raritan PX lineup can be found for reasonable prices. They have a few models that take standard 120v single-phase input and have standard NEMA 5-15R outlets. Let’s buy one, and see how it goes.

Over time, it seems that Xfinity moved regions to receive their DHCPv6 addresses from a server with a real address rather than a link-local address. The default firewall rules in OpenWRT have been fixed, but if you are running an older OpenWRT version, or you copied your firewall rules from an older version, you might need to fix this manually.

Find the rule called “Allow-DHCPv6” and remove any source and destination IP restrictions. Save and reload the firewall. That’s it.

Costco was practically giving away these ceiling fan + light combos. The problem is that they’re a complete PITA to control programmatically. They use a 433MHz remote. Fortunately, there is a way, it’s just not easy.

My homelab uses quite a bit of power – somewhere in the ballpark of 700w, but that includes general networking hardware as well. Combine that with a Threadripper workstation and a separate work PC setup, and the power meter is not too happy. So what can I do?

Intel’s RAPL technology allows you to measure the power usage of your CPU, including some sub-components such as the memory controller. It even works on some AMD CPUs. Measuring it requires some kind of scripting or automation, since the underlying measurements report the total microjoules consumed by the component since boot, rather than an instantaneous wattage. I have both a simple script with minimal dependencies, and a Zabbix template for RAPL measurements.

Server backplanes are a thing I care a little too much about. Despite them being a mainstay of most servers, there isn’t really a comprehensive guide about them. Now, there is.

If your system has the capability of monitoring ECC errors via /sys/devices/system/edac/mc*/ items, you can use this template to do so. I had found a template elsewhere on the internet, but it was hardcoded to only monitor mc0, the first memory controller. Many CPUs that support ECC have multiple memory controllers, not to mention multi-socket systems.

This error can occur if you are trying to perform a recovery from OSDs, and the cluster_uuid extracted from the recovery does not match the monmap.

# Replace 'pve1' with the name of your monitor
# Stop the monitor
systemctl stop ceph-mon@pve1
# Extract the monitor map to a file called monmap
ceph-mon -i pve1 --extract-monmap monmap
# Change FSID
monmaptool --clobber --fsid $NEW_FSID monmap
# Make any other changes via monmaptool, such as rewriting the monitor list
# Inject the new monmap
ceph-mon -i pve1 --inject-monmap monmap
# Start service
systemctl start ceph-mon@pve1