About

I’m Damien Dye.

Presales Engineer covering Europe and APAC at croit GmbH, a founding member of the Ceph Foundation and an official Proxmox Gold Partner.

Twenty-odd years of it has been the paid part: Microsoft platforms, Linux, enterprise applications, virtualisation, networking, storage and security. The paid part is not the whole of it. I’ve been building things and breaking them since the mid-nineties and running Linux properly from 1999, years before anybody thought it worth paying me for, and I count those years because you learn as much from kit that is yours as from kit somebody else has insured.

South Yorkshire, so you will get it straight. If summat works I’ll tell you why. If it doesn’t I’ll tell you that as well, and I would rather say so before you have spent the money than after.

The early years

I’ve been taking computers apart since I was eight. My first machine was an Atari STfm with 512K of RAM.

My first PC came at twelve, running Windows 3.11, and from there I went through the lot of them: 95, 95b, 95c, 98, 98SE and then Windows 2000.

That wasn’t using software. It was working out what changed between one version and the next, what broke on the way, and how to get a thing running when it had decided it would rather not.

My first internet connection was a 56k dial-up modem on Freeserve — one of the first free ISPs in the UK. I moved to ADSL as soon as it became available in 2001, through Demon Internet. Then to VDSL in 2008, and finally to FTTP with Zen Internet in 2017 — which is where native IPv6 arrived. That’s where it all started in terms of networking, and it’s a long way from the 100GbE fabrics I design now. But the curiosity was the same.

The local network started even earlier, and rougher. My first LAN was 10BASE2 — thin coaxial cable, BNC connectors and a 50-ohm terminator at each end, every machine sharing a single 10 Mbit bus. I got two PCs talking over it, then added a 5-port hub as more machines turned up. That hub later became a switch — a real step up, giving each port its own collision domain instead of everything fighting over shared coax. Wireless came next, as soon as it was affordable: 802.11b at 11 Mbit, on Orinoco Gold PCMCIA cards in the laptops. Coax bus to shared hub to switched Ethernet to Wi-Fi — I worked through every step of that by hand.

I also built fully unattended Windows XP installations. These used the old DriverPacks for driver injection and custom scripts for automated application installs. You could go from bare metal to a fully configured system without touching the keyboard. That was automation thinking years before I ever heard of Ansible — and it was on Windows, not Linux.

Finding Linux

I started with Linux at fifteen, on SuSE 6. I went straight for the distributions that made you understand what was happening underneath.

Over the Christmas holidays in 2001, I built a complete system using the Linux From Scratch book. Every package compiled by hand. Every dependency understood. Every configuration decision made deliberately.

By 2002 I’d moved to Gentoo. Gentoo runs on the same principle: you build the whole system from source, you understand what every USE flag does, and when something breaks you know exactly where to look.

Exploring other platforms

I never stuck to one architecture.

I had a DEC Alpha system from 2001 to 2007. The Alpha was Digital Equipment Corporation’s 64-bit RISC processor, launched in 1992 — a true 64-bit machine more than a decade before x86 caught up with AMD64 in 2003. It ran Tru64 UNIX, OpenVMS, Windows NT and Linux, and for a time it was about the fastest thing you could put on a desk. Through community discussions I got USB and FireWire working on it. I even fitted a PCMCIA-to-ISA adapter in the back, so the same PC Cards I used in the laptops — the Orinoco Gold wireless among them — would run in the desktop Alpha. Not a small thing on a platform where nowt was guaranteed to work.

I had a go at BeOS. It took a completely different run at multithreading and multimedia from anything else around at the time.

At university, a mate and I rescued Sun SPARC workstations that were being thrown out and built Gentoo on them. SPARC was Sun Microsystems’ RISC architecture, introduced in 1987 — the engine behind the Sun workstations and servers that ran much of the Unix world through the nineties, usually under Solaris. Getting a from-source Linux running on that hardware was the whole point. Because why wouldn’t you, if the kit’s there and you want to see if it works.

I’ve also built and worked with Linux on ARM and ARM64, using Raspberry Pis and Odroids. And I’ve run Windows on ARM64.

Run the same operating system across x86, Alpha, SPARC, ARM and ARM64 and the assumptions you made on one don’t follow you to the next — byte order, alignment, page sizes, driver support and toolchain quirks all shift underneath you. That matters more than ever now ARM64 sits in the data centre alongside x86, and it’s the thinking that keeps a mixed-architecture Ceph or Proxmox estate honest.

I was also experimenting with IPv6 early on. I had access to 6bone — the experimental IPv6 test network. 6bone was a global testbed that ran from 1996 to help develop and deploy IPv6 before the production internet was ready for it. It carried IPv6 mostly over IPv4 tunnels, used its own 3ffe::/16 address range, and was deliberately shut down on 6 June 2006 once native IPv6 had matured enough to stand on its own. I ran both the Linux IPv6 stack and the Microsoft Research IPv6 stack for Windows XP. I’ve been through the lot. Started with IPv6-in-IPv4 tunnels. Moved to 6to4 (RFC 3056) for automatic tunnelling. Then on to full native IPv6 when I moved to a decent ISP — Zen Internet. Most people didn’t touch IPv6 until their employer made them. I’d already been through every transition mechanism by that point, across both Linux and Windows, because I wanted to understand where networking was heading. That’s why IPv6 feels natural now rather than summat bolted on after the fact. Now, ninety percent of my traffic runs over native IPv6. I run a browser extension called IPvFoo to show me what each connection is using and whether sites are serving mixed protocols or IPv4 only. Old habits — I like to see what’s actually happening, not just assume.

Learning by doing

Everything I know technically I learnt by having a go at it. Built things, broke them, worked out why they broke, built them again.

The degree was Business Studies and Computer Network Engineering at Sheffield Hallam, and it gave me the one thing teaching yourself doesn’t — how a business actually works, how a technical decision turns into a commercial outcome, and how to think about a system in terms of the problem it solves for whoever is paying for it. That has shaped every role since.

The technical skills came from doing, though, not from studying. Production problems don’t arrive with a reading list. Being able to work a thing out from first principles is worth more than a certificate on the wall, and it does not expire.

Where open source comes from

When Microsoft’s leadership described open source as “a cancer” in the early 2000s, I was already deep into Linux. Building systems from scratch. Running Gentoo. Contributing to communities.

That sort of hostility toward people sharing knowledge and building things together didn’t put me off. It pushed me further in. If a company’s answer to collaborative development is to call it a disease, that tells you more about the company than the software. I dug deeper into open source and made it my go-to rather than backing that mindset.

Over the years, the practical case caught up with the principled one. Proprietary platforms work well enough until the vendor changes the licensing, gets bought, or decides your use case isn’t worth keeping. Then you’re stuck. Your data, your workflows and your team’s know-how are all tied to a platform you no longer control. The Broadcom acquisition of VMware is the most recent and visible example, but it’s far from the only one.

Same thinking applies to cloud. Ask what you are actually renting and the answer is capacity you could have owned outright, on a meter that never stops running. The promised benefits — agility and elasticity and less to manage — rarely land in the shape the pitch described, and the cost builds year on year in a way that owned kit does not. As such I have been able to show, every time anybody has asked me to, that open source on well-designed hardware gives better value, more control and fewer surprises. Not a fashionable position. It is one I can put numbers against.

Building a career

My career started in 2005 at a precision castings manufacturer in Worksop. IT engineer, part of a small team, supporting around fifty users.

That first role covered a proper breadth. Maintaining the Manusoft ERP system and document archiving. Linux and Windows 2003 server administration. Crystal Reports development for production and maintenance decision-making. PBX administration. CAD/CAM system integration and management of computer-aided robotic manufacturing. Hardware and software procurement. Training end users and handling support at their workstations.

I was also writing automation from day one. I upgraded the company’s Active Directory and wrote VB scripts to automatically map drives, assign printers, and deploy software based on group membership. That was 2006 — proper infrastructure automation in my first professional role.

That’s manufacturing IT. When the line stops because of something you look after, you find out quickly that reliability is not a preference, and the people stood next to the machine will tell you so themselves. It was also the first time I ran Linux and Windows in the same building for a living, which has been the pattern ever since.

From there I moved into front-line technical support. First and second line, working on a dedicated desk for a large multinational customer. Windows desktops, Active Directory, Exchange 2007, Cisco VPN, Cisco Call Manager, ITIL-based ticket management on Remedy. That ITIL discipline — change management, problem management, structured processes — followed me into every role since. I also bridged the Linux and Windows worlds early — configuring Services for Unix and resolving Citrix access to Unix NFS shares.

Even on that desk I was building past the job description. I wrote an interface that provisioned and disabled user accounts straight off the HR data in Agresso, handling the Active Directory account, the group memberships, the Office Communicator configuration, the Exchange mailbox and the aliases. That is systems integration from a first-line seat, which was not what the job title said.

One of the customers I supported on that desk later hired me at the next company.

Next was a global comms company, ICT and security. That’s where the application development properly started. Over two and a half years I built or contributed to six distinct internal systems. A custom web-based quoting tool. Order processing interfaces between Salesforce and Agresso. A SharePoint master data catalogue fed from Salesforce and Agresso. A custom web project management system with interfaces to MSPE and Agresso for financial processing. And a migration from a customised Salesforce Service Cloud application to ServiceNow with full ITIL process implementation. I was also writing SQL interfacing architecture for corporate business applications and tuning MS SQL performance. I administered IIS and Windows Server 2008 Terminal Services, and owned the corporate data model. Directory services — Active Directory and LDAP — became a core skill here that followed me through every role afterwards.

I also ran a UK-wide Windows 7 replacement programme, moving the entire UK userbase onto new HP kit in three weeks, with 95% of them reporting no disruption to their work at all. Three weeks is the sort of number that only comes off if the preparation was done properly, and the preparation is the unglamorous half nobody asks about afterwards.

Then came management. A semiconductor design house with a small team of engineers spread across several countries. That pulled a few threads together at once. I was developing on the Force.com platform — triggers, Visualforce pages, custom controllers — to integrate Financial Force accounting and PSA systems globally. At the same time I was building two separate PXE boot HPC clusters. One for the UK European engineering operations and one for the Chinese engineering operations. Both used Red Hat with NFS root filesystems to create consistent, 100% repeatable diskless compute farms. I deployed ZFS on Linux with Dell PowerVault equipment for unified storage. I replaced legacy isolated security environments with a Windows Server Active Directory using native Kerberos and LDAP for single sign-on across Linux and Windows. I stabilised international connectivity by building a unified network using OpenVPN tunnelling across difficult operating conditions into China.

For a fair chunk of that role I was the only person covering all of it: the Linux compute farms, the Windows support, the Salesforce development and user support across several time zones, under SoC design workloads on advanced technology nodes. One of the engineers who reported to me went on to be a Salesforce developer, and I count that as the better outcome of those years.

Next was a proper deep dive into Linux. Third-line support at Pulsant, a cloud and hosting provider, working across the full stack. RHEL, CentOS, Ubuntu. MySQL clustering with Galera, MongoDB sharding, HAProxy with SNI and SSL termination, Apache, PostgreSQL, PHP-FPM tuning for high-performance e-commerce, Postfix mail, BIND and PowerDNS for DNS hosting, Varnish for web caching, Squid for proxy caching. IPTables, IPset, and Cisco ASA for firewalling and DoS protection. Linux server performance tuning for network and storage. CPanel troubleshooting, SolarWinds monitoring template development, and New Relic administration. All running on VMware 5.5 with vCloud Director underneath.

It wasn’t only support either. I designed a Galera database replication product there and took it from concept through prototyping to a service customers paid for. That wasn’t a lab exercise. It was built and proved against real workloads belonging to real paying customers, which is not a thing a hosting provider hands out lightly.

Third line teaches you what being the last line of escalation means. When it reaches you, nobody behind you is going to fix it.

Then Nominet — the registry behind every .uk domain name. DNS at national scale. The infrastructure has to be reight solid twenty-four hours a day, every day, with zero downtime and out-of-hours on-call. VMware 5.5 and 6, HP 3par storage with fibre-channel zoning on Brocade, RHEL 6 and 7 managed with Puppet, F5 load balancers across IPv4 and IPv6, Postfix mail, and a Zabbix deployment I built to replace the ageing VMware Hyperic monitoring. I also adopted and customised ServiceNow, implementing workflows, application deployment, and Linux node discovery for configuration management and inventory. I helped design and build processes for out-of-hours service desk outsourcing to reduce on-call burden.

At Nominet I ended up being the first person asked, whether it was Linux, Unix, ServiceNow or something nobody could place, and I made a point of coming back with something that worked rather than something that sounded good. That is the sort of place where you learn that the boring, disciplined approach to infrastructure is the one that survives contact with a Tuesday morning.

After Nominet I moved across infrastructure and hosting. VMware 6.7, Zerto for disaster recovery, Dell Compellent and Nexsan storage with Brocade fibre-channel zoning, Citrix Cloud with FSLogix profile management alongside Azure. I introduced Zabbix monitoring there too, designing the templates and scripts from scratch.

Then retail. Line managing a small team, pulling VMware 6.7 admin back in-house from a third party, deploying Zabbix monitoring (again, replacing somebody’s failed attempt), rebuilding OS deployment around PXE and Chocolatey, upgrading the network with Fortinet kit, and sorting out hardware procurement.

Then the role that changed everything. At the UK Centre for Ecology & Hydrology, I managed the science computing team — four direct reports — and rebuilt the infrastructure from the ground up. A 7-node Proxmox VE private cloud with hyperconverged Ceph storage on dual 100Gb switching. An 8-node HPC cluster on HDR InfiniBand running Slurm, with SR-IOV for VM access to the fabric and EasyBuild for software builds. Migration from GPFS to all-NVMe Ceph storage. Ansible with Netbox as the source of truth for everything — patch management, configuration drift control, integrations into Cloudflare, PowerDNS, and Active Directory. OSPF dynamic routing for cloud networks. Local repository mirrors for maximum deployment speed and consistency. CentOS 7 to Rocky 9 redeployment of the HPC environment. Cloudflare for DNS (including DNSSEC), DDoS protection, and Zero Trust remote access. This included migrating 25 authoritative DNS zones from self-hosted BIND 9 to Cloudflare in two working days. Multiple registries were updated, and the whole thing was integrated with Ansible and Let’s Encrypt.

All of it on open-source tools, tight budget, built deliberately to stay clear of the Broadcom and VMware licensing trap before it shut. That settled the argument. Open-source infrastructure at that scale is not just workable — it is better, and I have the cluster and the invoices to say so.

The other half of that job was the four people in the team. Scientists do not care what the storage is called. They care whether the job runs tonight, and whether the person they ask can explain the answer without making them feel daft for asking.

How it all comes together

The move into presales at croit wasn’t a change of direction. It was everything landing in one place.

Manufacturing IT taught me that reliability stops being a preference the moment production depends on it. Front-line support taught me to listen before doing anything else. Application development taught me to build systems that solve a business problem rather than an interesting one. Managing a global team taught me to carry the technical load and still develop the people around me, which is harder than either half on its own. Third-line Linux taught me what the last line of escalation feels like. Running Windows and Linux side by side for twenty years taught me how platforms actually behave under load, as against how the datasheet says they will. VMware, which I ran in nearly every role from 2008 on, taught me what enterprise virtualisation looks like at scale — and then what happens when the commercial ground shifts underneath a platform the whole estate is sat on. Storage and networking taught me where the hard problems live. Security has run through the lot, from firewalls and VPNs early on to DNSSEC, Zero Trust and hardening now. Nominet taught me discipline.

Put all that together and you get presales. You design the system, then you defend the design, and you stay honest about what it will not do, because somebody is about to spend real money on your say-so.

At croit, that means working with outfits across Europe and Asia-Pacific who are rethinking their virtualisation and storage. The conversations are usually about leaving VMware for Proxmox VE with Ceph. The workloads coming across are mostly Windows, so the cross-platform experience isn’t old news. It’s what I do now.

The part I care about is the honesty. What I put in front of somebody has to be the thing that works in their building, at their scale, with the constraints they actually have, and if their existing kit does most of the job already then that is what I will tell them.

Storage

Storage kept being where the hardest problems sat. It wasn’t a deliberate career plan — it just worked out that way.

The fascination started young. In the 90s I dreamed of Iomega Zip and Jaz disks — 100MB, and then a whole gigabyte, on a single removable cartridge, when the floppies everyone else swapped around held just 1.44MB. They were expensive kit, so for years they stayed a want rather than a have. By the time I finally got a USB Zip drive, USB flash drives had just started to appear — and the format I’d wanted for so long was already on its way out. An early lesson in how fast storage moves, and how quickly today’s must-have becomes tomorrow’s drawer clutter.

Optical was part of the same story. In 1998 I had a quad-speed HP CD-RW drive, and it was a cracker. It could read and rewrite discs that later, faster drives simply refused — so it earned its keep as a rescue drive long after it should have been retired.

It started at the interconnect. I’ve worked with storage hardware across the lot. IDE, multiple generations of SCSI, SATA, SAS, and NVMe on the direct-attached side. ATA over Ethernet and iSCSI on the network storage side. HP 3par, Dell Compellent, Dell PowerVault, Nexsan — each with its own quirks and failure modes.

From there it went up through the stack. Clustered LVM taught me how shared storage behaves when multiple nodes need concurrent access — and what happens when locking and fencing aren’t right. ZFS taught me what happens when you actually think about data integrity at the filesystem level. GPFS showed me parallel filesystems at scale. Ceph taught me how distributed systems behave differently under failure.

Over the years, what started as “the person who also does the storage” became “the person you call when the storage needs to be designed properly.”

Networking

Networking has been there from the start. It’s not a supporting skill — it’s a core one. TCP/IP, DHCP, and DNS span every single role I’ve held from McKenna onwards.

Working at Nominet meant working on the infrastructure behind the UK’s domain name registry. That’s DNS at national scale. Naming, resolution, delegation, and the expectation that it works every single time.

I ran authoritative zones on BIND 9, configured DNSSEC, set up F5 load balancers for IPv4 and IPv6, and later migrated DNS estates to Cloudflare with API automation and Let’s Encrypt integration. I’ve built PXE boot DNS infrastructure for cluster deployments in both the UK and China. I understand DNS from both sides. Self-hosted, where you own every failure. And managed, where you’re trusting a provider and need to verify that trust through monitoring.

Beyond DNS, networking runs through every role I’ve held. VLANs, bonding, LACP, fibre-channel zoning with Brocade, Fortinet and Cisco ASA firewalling, IPTables and IPset. 25GbE and 100GbE fabric design, BGP, OSPF dynamic routing, MTU management, IPv6 architecture. VPN tunnelling — from OpenVPN across difficult international links to WireGuard and cloudflared for modern secure access. Samba has been a professional skill across multiple roles too, bridging Linux file sharing and Windows domain integration long before I was testing the AD implementation in alpha.

A Ceph cluster is a network application. The storage performance ceiling is set by the network underneath it. The failure modes are network failure modes. That’s not summat you learn from a textbook. It’s summat you learn from troubleshooting at two in the morning.

Linux

I’ve worked across RHEL, Debian, SUSE, and Fedora families for over two decades. Professionally, that’s meant RHEL and CentOS for live services, Ubuntu for application hosting, Rocky for HPC, and Gentoo and Fedora for personal use. I’ve passed LinkedIn’s Linux skill assessment.

The depth came from problems that have no Stack Overflow answer, which is where you end up learning the PCI subsystem properly rather than by rumour: IOMMU groups, ACS, VFIO, SR-IOV, Resizable BAR and what DMA translation is quietly costing you, and kernel boot parameters as controls that change how the machine behaves rather than a list to copy off a wiki because somebody’s blog said it fixed their thing.

Storage and virtualisation performance problems nearly always trace back to a layer nobody looked at. That is where my Linux knowledge lives.

Windows

Linux is where I spend my time now, but the Windows side is just as real and it has been there in every job I have had. Windows Server, Active Directory, LDAP, Exchange, IIS, Microsoft SQL Server. None of that is a legacy skill I have put down.

It matters at the guest level, which is where most people stop looking. When a Windows workload is running on KVM, the person who can tell you how that guest behaves under a given CPU topology, and can trace a performance complaint back to a Microsoft patch instead of blaming the hypervisor, is the person who has spent years on both sides of the fence. Most VM performance arguments I have watched were won by somebody knowing the guest, not the host.

The people I have supported run from professors and board directors to the team that looks after the buildings. The job is the same in every case. Find out what they actually need, get it working, and explain it in a way that doesn’t leave them feeling daft for having asked.

VMware

VMware is the platform I grew up on professionally, across seven roles from 2008 onwards, and I ran the full stack of it: ESXi, vCenter, vSAN, vSphere clustering, vMotion. I know what it does well. I know where it doesn’t. And I watched the Broadcom acquisition rewrite the commercial reality underneath outfits that had built the whole estate on it, which is a different thing from reading about it.

You cannot help somebody leave a platform you never properly worked on. As such I can tell them what they are giving up, what they are getting, and which part of the migration will be worse than they have been led to expect.

Automation

Automation thinking started in my first professional role in 2006. VB scripts at McKenna to automate AD drive mapping, printer assignment, and software deployment. Then HR-to-AD provisioning tools at BT Engage IT. Then PXE boot systems for diskless compute clusters at Sondrel. Then Puppet at Nominet. Then Chocolatey packaging and PXE rebuild at a retail company. Then ServiceNow workflow customisation across multiple roles.

Now it’s Ansible with Netbox as the source of truth. Not because they’re trendy, but because infrastructure that can’t be rebuilt from code isn’t infrastructure you can trust.

Documentation is the same argument. Knowledge that lives only in somebody’s head is a single point of failure, and it walks out of the building at five o’clock with the rest of them. Same as a disk with no redundancy behind it, and treated with the same seriousness.

Monitoring

My monitoring background started with SolarWinds at Pulsant, running scale monitoring across the hosting estate. Zabbix came later, and it has earned its own mention because I’ve put it in at nearly every place since. It began at Nominet, where I ran the tender and the testing and then moved us onto Zabbix to replace the ageing VMware Hyperic setup, and it won on being genuinely intuitive rather than another thing wearing Nagios underneath. After that I put it in from scratch at a hosting and infrastructure platform, replaced somebody’s failed attempt at it in retail, and used it to watch an 8-node HPC cluster at UKCEH.

Each time I designed the monitoring templates and wrote the scripts myself. By the time I sat the Zabbix Specialist and Professional exams in 2018, I’d already been deploying it for years.

Communication

Working in critical infrastructure teaches you to be precise. Working in presales teaches you to be clear. The two are related but not the same.

Being precise means nowt if the person listening cannot follow the reasoning, so I have had to learn to do the same explanation twice over: once for the engineer who wants to see the CRUSH map, and once for whoever has to sign off why the budget is the number it is. I don’t dumb owt down. I just make every step in the reasoning visible, and let them stop me where they want to.

Years of end users taught me a different sort of patience. The people who reckon themselves the smartest in the room are usually the ones who behave the most helplessly. The ones who open with “I’m not very technical” tend to listen, follow the steps and get it sorted in ten minutes. And the ones who tell you they know exactly what they are doing have generally made it worse before they picked the phone up.

The business systems background helps here more than people expect, because I have always had to translate in both directions. SQL architecture for an operations manager, a storage design for a CTO, or sitting next to somebody at their own desk showing them the thing they were never shown. Same skill every time.

Community

Sharing knowledge has run through the whole career.

I was active on the Gentoo forums when I was building systems from source and needed to understand why a USE flag combination was breaking a compile. I contributed on the Samba forums when I was working through file sharing and domain integration problems across Linux and Windows. That went beyond just asking questions. I built a Samba-based Active Directory domain controller while the AD support was still in alpha. I tested it against Windows 2000, XP, and Vista clients and fed back into the community.

Now I’m an active contributor on the Proxmox community forums as DamienDye. I help with NVMe passthrough performance, Windows VM tuning, Ceph troubleshooting, and cluster networking.

The technologies change. The principle doesn’t. If I have worked a problem out, there is nowt to be gained by sitting on it, and somebody at two in the morning six months from now will be glad it is written down.

I also have a habit of digging into problems beyond what the task strictly needs. I’ve designed a DIY NVMe power loss protection board because I wanted to understand exactly why FTL corruption happens at the hardware level. I’ve researched self-hosted email protocols because I wanted to understand JMAP from the RFC up rather than just trusting a provider. I built this blog because writing things down properly is how you find the gaps in your own understanding.

Community off the keyboard

Not all of it has been forums.

In 2018 I was one of the founders of the Longford Park Community Association in Banbury, on the estate where I live. Four phases of new build, a community centre written into the plans, and nowt in existence to run it. So a handful of residents set summat up.

I did the IT end first, because that is what I had to give. The lpca.org.uk domain went live in January 2018, and behind it I built the website, the mail systems and the committee lists, and wrote the privacy policy.

I was on the committee from the start, and I was its chairman from December 2018 to February 2020. Fourteen months, and the things that actually mattered all landed inside them. We got it registered as a charity on 11 February 2019, number 1181953. I sorted the commercial lease for the building. Then we opened the centre.

That is what the chair was for. Not agendas and minutes. Getting a building open that a few hundred households could walk into. Resident volunteers still run it across phases 1 to 4 of the development — three rooms, a kitchen and a car park, hired out by the hour to whoever on the estate wants them.

Volunteer committees run on goodwill, and goodwill is not a governance model. So I held people to the constitution, myself included. I asked why we were recruiting from outside when the constitution said engage the residents, and I pulled up a mailout that was going to land in every spam folder on the estate. Not to be awkward. A residents’ association that cannot reach the residents has already failed at the only job it has, and I sent the links to fix it along with the complaint.

The centre is still open and I am not on the committee, and that is the right way round. Anything that only works while you are stood there holding it up was never built properly.

This site

This blog is a Hugo static site using the PaperMod theme. It’s hosted on Cloudflare Workers, serving the built site as static assets.

I write here about the infrastructure I work with day to day: Proxmox VE, Ceph, Ansible, Netbox, certificates and whatever else I have been digging into that week. It is written to be used, with the commands and the numbers in it, because a post you cannot follow at the keyboard is decoration. No marketing fluff. If summat has rough edges, the post says so.

Get in touch

You can find me on LinkedIn or on the Proxmox forums.

If you are looking at Ceph or Proxmox for your place and want a proper conversation about it rather than a pitch, drop me a line. Bring the workload, the constraints and the budget you actually have. I will tell you what it will do, what it won’t, and if the honest answer is that you should keep what you have got and configure it properly, you will get that answer as well.